SHARE
TWEET

Eng Do You Trust This Computer? (2018)

tpbolon Feb 4th, 2019 1,099 Never
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
  1. 1
  2. 00:00:15,580 --> 00:00:25,547
  3. ♪♪
  4.  
  5. 2
  6. 00:00:25,590 --> 00:00:35,035
  7. ♪♪
  8.  
  9. 3
  10. 00:00:35,078 --> 00:00:41,302
  11. What we're on the brink of is
  12. a world of increasingly intense,
  13.  
  14. 4
  15. 00:00:41,345 --> 00:00:45,219
  16. sophisticated
  17. artificial intelligence.
  18.  
  19. 5
  20. 00:00:45,262 --> 00:00:48,396
  21. Man: Technology is evolving
  22. so much faster than our society
  23.  
  24. 6
  25. 00:00:48,439 --> 00:00:51,181
  26. has the ability
  27. to protect us as citizens.
  28.  
  29. 7
  30. 00:00:51,486 --> 00:00:55,707
  31. The robots are coming, and they
  32. will destroy our livelihoods.
  33.  
  34. 8
  35. 00:00:55,751 --> 00:01:01,844
  36. ♪♪
  37.  
  38. 9
  39. 00:01:01,887 --> 00:01:04,238
  40. You have a networked
  41. intelligence that watches us,
  42.  
  43. 10
  44. 00:01:04,281 --> 00:01:08,590
  45. knows everything about us,
  46. and begins to try to change us.
  47.  
  48. 11
  49. 00:01:08,633 --> 00:01:12,768
  50. Man #2: Twitter has become the
  51. world's number-one news site.
  52.  
  53. 12
  54. 00:01:12,811 --> 00:01:15,205
  55. Man #3:
  56. Technology is never good or bad.
  57.  
  58. 13
  59. 00:01:15,249 --> 00:01:18,948
  60. It's what we do
  61. with the technology.
  62.  
  63. 14
  64. 00:01:18,991 --> 00:01:22,734
  65. Eventually, millions of people
  66. are gonna be thrown out of jobs
  67.  
  68. 15
  69. 00:01:22,778 --> 00:01:25,737
  70. because their skills
  71. are going to be obsolete.
  72.  
  73. 16
  74. 00:01:25,781 --> 00:01:27,435
  75. Woman: Mass unemployment...
  76.  
  77. 17
  78. 00:01:27,478 --> 00:01:32,527
  79. greater inequalities,
  80. even social unrest.
  81.  
  82. 18
  83. 00:01:32,570 --> 00:01:35,530
  84. Man #4: Regardless of whether
  85. to be afraid or not afraid,
  86.  
  87. 19
  88. 00:01:35,573 --> 00:01:38,185
  89. the change is coming,
  90. and nobody can stop it.
  91.  
  92. 20
  93. 00:01:38,228 --> 00:01:44,539
  94. ♪♪
  95.  
  96. 21
  97. 00:01:44,582 --> 00:01:46,323
  98. Man #5: We've invested
  99. huge amounts of money,
  100.  
  101. 22
  102. 00:01:46,367 --> 00:01:49,283
  103. and so it stands to reason
  104. that the military,
  105.  
  106. 23
  107. 00:01:49,326 --> 00:01:50,893
  108. with their own desires,
  109.  
  110. 24
  111. 00:01:50,936 --> 00:01:53,330
  112. are gonna start to use
  113. these technologies.
  114.  
  115. 25
  116. 00:01:53,374 --> 00:01:55,419
  117. Man #6:
  118. Autonomous weapons systems
  119.  
  120. 26
  121. 00:01:55,463 --> 00:01:57,552
  122. could lead to a global arms race
  123.  
  124. 27
  125. 00:01:57,595 --> 00:02:00,032
  126. to rival the Nuclear Era.
  127.  
  128. 28
  129. 00:02:00,076 --> 00:02:02,296
  130. ♪♪
  131.  
  132. 29
  133. 00:02:02,339 --> 00:02:04,036
  134. Man #7:
  135. We know what the answer is.
  136.  
  137. 30
  138. 00:02:04,080 --> 00:02:05,429
  139. They'll eventually
  140. be killing us.
  141.  
  142. 31
  143. 00:02:05,473 --> 00:02:10,782
  144. ♪♪
  145.  
  146. 32
  147. 00:02:10,826 --> 00:02:12,349
  148. Man #8:
  149. These technology leaps
  150.  
  151. 33
  152. 00:02:12,393 --> 00:02:15,874
  153. are gonna yield
  154. incredible miracles...
  155.  
  156. 34
  157. 00:02:15,918 --> 00:02:18,181
  158. and incredible horrors.
  159.  
  160. 35
  161. 00:02:18,225 --> 00:02:24,231
  162. ♪♪
  163.  
  164. 36
  165. 00:02:24,274 --> 00:02:29,323
  166. Man #9: We created it,
  167. so I think, as we move forward,
  168.  
  169. 37
  170. 00:02:29,366 --> 00:02:33,762
  171. this intelligence
  172. will contain parts of us.
  173.  
  174. 38
  175. 00:02:33,805 --> 00:02:35,981
  176. And I think the question is --
  177.  
  178. 39
  179. 00:02:36,025 --> 00:02:39,463
  180. Will it contain
  181. the good parts...
  182.  
  183. 40
  184. 00:02:39,507 --> 00:02:41,378
  185. or the bad parts?
  186.  
  187. 41
  188. 00:02:41,422 --> 00:02:47,079
  189. ♪♪
  190.  
  191. 42
  192. 00:02:57,742 --> 00:03:04,793
  193. ♪♪
  194.  
  195. 43
  196. 00:03:04,836 --> 00:03:08,840
  197. Sarah: The survivors
  198. called the war "Judgment Day."
  199.  
  200. 44
  201. 00:03:08,884 --> 00:03:12,583
  202. They lived only to face
  203. a new nightmare --
  204.  
  205. 45
  206. 00:03:12,627 --> 00:03:14,019
  207. the war against the machines.
  208.  
  209. 46
  210. 00:03:14,063 --> 00:03:15,412
  211. Aah!
  212.  
  213. 47
  214. 00:03:15,456 --> 00:03:18,023
  215. Nolan: I think
  216. we've completely fucked this up.
  217.  
  218. 48
  219. 00:03:18,067 --> 00:03:21,549
  220. I think Hollywood has managed
  221. to inoculate the general public
  222.  
  223. 49
  224. 00:03:21,592 --> 00:03:24,247
  225. against this question --
  226.  
  227. 50
  228. 00:03:24,291 --> 00:03:28,251
  229. the idea of machines
  230. that will take over the world.
  231.  
  232. 51
  233. 00:03:28,295 --> 00:03:30,645
  234. Open the pod bay doors, HAL.
  235.  
  236. 52
  237. 00:03:30,688 --> 00:03:33,561
  238. I'm sorry, Dave.
  239.  
  240. 53
  241. 00:03:33,604 --> 00:03:35,911
  242. I'm afraid I can't do that.
  243.  
  244. 54
  245. 00:03:37,434 --> 00:03:38,696
  246. HAL?
  247.  
  248. 55
  249. 00:03:38,740 --> 00:03:40,437
  250. Nolan:
  251. We've cried wolf enough times...
  252.  
  253. 56
  254. 00:03:40,481 --> 00:03:42,483
  255. HAL?
  256. ...that the public
  257. has stopped paying attention,
  258.  
  259. 57
  260. 00:03:42,526 --> 00:03:43,962
  261. because it feels like
  262. science fiction.
  263.  
  264. 58
  265. 00:03:44,006 --> 00:03:45,486
  266. Even sitting here talking
  267. about it right now,
  268.  
  269. 59
  270. 00:03:45,529 --> 00:03:48,228
  271. it feels a little bit silly,
  272. a little bit like,
  273.  
  274. 60
  275. 00:03:48,271 --> 00:03:51,666
  276. "Oh, this is an artifact
  277. of some cheeseball movie."
  278.  
  279. 61
  280. 00:03:51,709 --> 00:03:56,584
  281. The WOPR spends all its time
  282. thinking about World War III.
  283.  
  284. 62
  285. 00:03:56,627 --> 00:03:59,064
  286. But it's not.
  287.  
  288. 63
  289. 00:03:59,108 --> 00:04:02,111
  290. The general public is about
  291. to get blindsided by this.
  292.  
  293. 64
  294. 00:04:02,154 --> 00:04:11,512
  295. ♪♪
  296.  
  297. 65
  298. 00:04:11,555 --> 00:04:13,514
  299. As a society and as individuals,
  300.  
  301. 66
  302. 00:04:13,557 --> 00:04:18,954
  303. we're increasingly surrounded
  304. by machine intelligence.
  305.  
  306. 67
  307. 00:04:18,997 --> 00:04:22,653
  308. We carry this pocket device
  309. in the palm of our hand
  310.  
  311. 68
  312. 00:04:22,697 --> 00:04:24,829
  313. that we use to make
  314. a striking array
  315.  
  316. 69
  317. 00:04:24,873 --> 00:04:26,831
  318. of life decisions right now,
  319.  
  320. 70
  321. 00:04:26,875 --> 00:04:29,007
  322. aided by a set
  323. of distant algorithms
  324.  
  325. 71
  326. 00:04:29,051 --> 00:04:30,748
  327. that we have no understanding.
  328.  
  329. 72
  330. 00:04:30,792 --> 00:04:34,143
  331. ♪♪
  332.  
  333. 73
  334. 00:04:34,186 --> 00:04:36,537
  335. We're already pretty jaded
  336. about the idea
  337.  
  338. 74
  339. 00:04:36,580 --> 00:04:37,929
  340. that we can talk to our phone,
  341.  
  342. 75
  343. 00:04:37,973 --> 00:04:40,062
  344. and it mostly understands us.
  345.  
  346. 76
  347. 00:04:40,105 --> 00:04:42,456
  348. Woman: I found quite a number
  349. of action films.
  350.  
  351. 77
  352. 00:04:42,499 --> 00:04:44,327
  353. Five years ago -- no way.
  354.  
  355. 78
  356. 00:04:44,371 --> 00:04:47,678
  357. Markoff: Robotics.
  358. Machines that see and speak...
  359.  
  360. 79
  361. 00:04:47,722 --> 00:04:48,897
  362. Woman: Hi, there....and listen.
  363.  
  364. 80
  365. 00:04:48,940 --> 00:04:50,202
  366. All that's real now.
  367.  
  368. 81
  369. 00:04:50,246 --> 00:04:51,639
  370. And these technologies
  371.  
  372. 82
  373. 00:04:51,682 --> 00:04:55,686
  374. are gonna fundamentally
  375. change our society.
  376.  
  377. 83
  378. 00:04:55,730 --> 00:05:00,212
  379. Thrun: Now we have this great
  380. movement of self-driving cars.
  381.  
  382. 84
  383. 00:05:00,256 --> 00:05:01,953
  384. Driving a car autonomously
  385.  
  386. 85
  387. 00:05:01,997 --> 00:05:06,088
  388. can move people's lives
  389. into a better place.
  390.  
  391. 86
  392. 00:05:06,131 --> 00:05:07,916
  393. Horvitz: I've lost
  394. a number of family members,
  395.  
  396. 87
  397. 00:05:07,959 --> 00:05:09,570
  398. including my mother,
  399.  
  400. 88
  401. 00:05:09,613 --> 00:05:11,876
  402. my brother and sister-in-law
  403. and their kids,
  404.  
  405. 89
  406. 00:05:11,920 --> 00:05:14,009
  407. to automobile accidents.
  408.  
  409. 90
  410. 00:05:14,052 --> 00:05:18,405
  411. It's pretty clear we could
  412. almost eliminate car accidents
  413.  
  414. 91
  415. 00:05:18,448 --> 00:05:20,102
  416. with automation.
  417.  
  418. 92
  419. 00:05:20,145 --> 00:05:21,843
  420. 30,000 lives in the U.S. alone.
  421.  
  422. 93
  423. 00:05:21,886 --> 00:05:25,455
  424. About a million around the world
  425. per year.
  426.  
  427. 94
  428. 00:05:25,499 --> 00:05:27,501
  429. Ferrucci:
  430. In healthcare, early indicators
  431.  
  432. 95
  433. 00:05:27,544 --> 00:05:29,503
  434. are the name of the game
  435. in that space,
  436.  
  437. 96
  438. 00:05:29,546 --> 00:05:33,158
  439. so that's another place where
  440. it can save somebody's life.
  441.  
  442. 97
  443. 00:05:33,202 --> 00:05:35,726
  444. Dr. Herman: Here in
  445. the breast-cancer center,
  446.  
  447. 98
  448. 00:05:35,770 --> 00:05:38,381
  449. all the things that
  450. the radiologist's brain
  451.  
  452. 99
  453. 00:05:38,425 --> 00:05:43,386
  454. does in two minutes, the
  455. computer does instantaneously.
  456.  
  457. 100
  458. 00:05:43,430 --> 00:05:47,303
  459. The computer has looked
  460. at 1 billion mammograms,
  461.  
  462. 101
  463. 00:05:47,347 --> 00:05:49,261
  464. and it takes that data
  465. and applies it
  466.  
  467. 102
  468. 00:05:49,305 --> 00:05:51,438
  469. to this image instantaneously,
  470.  
  471. 103
  472. 00:05:51,481 --> 00:05:54,441
  473. so the medical application
  474. is profound.
  475.  
  476. 104
  477. 00:05:56,399 --> 00:05:57,705
  478. Zilis:
  479. Another really exciting area
  480.  
  481. 105
  482. 00:05:57,748 --> 00:05:59,402
  483. that we're seeing
  484. a lot of development in
  485.  
  486. 106
  487. 00:05:59,446 --> 00:06:03,275
  488. is actually understanding
  489. our genetic code
  490.  
  491. 107
  492. 00:06:03,319 --> 00:06:06,104
  493. and using that
  494. to both diagnose disease
  495.  
  496. 108
  497. 00:06:06,148 --> 00:06:07,758
  498. and create
  499. personalized treatments.
  500.  
  501. 109
  502. 00:06:07,802 --> 00:06:11,632
  503. ♪♪
  504.  
  505. 110
  506. 00:06:11,675 --> 00:06:14,112
  507. Kurzweil:
  508. The primary application
  509. of all these machines
  510.  
  511. 111
  512. 00:06:14,156 --> 00:06:17,246
  513. will be to extend
  514. our own intelligence.
  515.  
  516. 112
  517. 00:06:17,289 --> 00:06:19,422
  518. We'll be able to make
  519. ourselves smarter,
  520.  
  521. 113
  522. 00:06:19,466 --> 00:06:22,643
  523. and we'll be better
  524. at solving problems.
  525.  
  526. 114
  527. 00:06:22,686 --> 00:06:24,775
  528. We don't have to age.
  529. We'll actually understand aging.
  530.  
  531. 115
  532. 00:06:24,819 --> 00:06:27,125
  533. We'll be able to stop it.
  534.  
  535. 116
  536. 00:06:27,169 --> 00:06:29,519
  537. Man: There's really no limit
  538. to what intelligent machines
  539.  
  540. 117
  541. 00:06:29,563 --> 00:06:30,868
  542. can do for the human race.
  543.  
  544. 118
  545. 00:06:30,912 --> 00:06:36,265
  546. ♪♪
  547.  
  548. 119
  549. 00:06:36,308 --> 00:06:39,399
  550. How could a smarter machine
  551. not be a better machine?
  552.  
  553. 120
  554. 00:06:42,053 --> 00:06:44,708
  555. It's hard to say exactly
  556. when I began to think
  557.  
  558. 121
  559. 00:06:44,752 --> 00:06:46,971
  560. that that was a bit naive.
  561.  
  562. 122
  563. 00:06:47,015 --> 00:06:56,459
  564. ♪♪
  565.  
  566. 123
  567. 00:06:56,503 --> 00:06:59,288
  568. Stuart Russell,
  569. he's basically a god
  570.  
  571. 124
  572. 00:06:59,331 --> 00:07:00,898
  573. in the field
  574. of artificial intelligence.
  575.  
  576. 125
  577. 00:07:00,942 --> 00:07:04,380
  578. He wrote the book that almost
  579. every university uses.
  580.  
  581. 126
  582. 00:07:04,424 --> 00:07:06,948
  583. Russell: I used to say it's the
  584. best-selling AI textbook.
  585.  
  586. 127
  587. 00:07:06,991 --> 00:07:10,255
  588. Now I just say "It's the PDF
  589. that's stolen most often."
  590.  
  591. 128
  592. 00:07:10,299 --> 00:07:13,650
  593. ♪♪
  594.  
  595. 129
  596. 00:07:13,694 --> 00:07:17,306
  597. Artificial intelligence is
  598. about making computers smart,
  599.  
  600. 130
  601. 00:07:17,349 --> 00:07:19,830
  602. and from the point
  603. of view of the public,
  604.  
  605. 131
  606. 00:07:19,874 --> 00:07:21,484
  607. what counts as AI
  608. is just something
  609.  
  610. 132
  611. 00:07:21,528 --> 00:07:23,268
  612. that's surprisingly intelligent
  613.  
  614. 133
  615. 00:07:23,312 --> 00:07:25,488
  616. compared to what
  617. we thought computers
  618.  
  619. 134
  620. 00:07:25,532 --> 00:07:28,404
  621. would typically be able to do.
  622.  
  623. 135
  624. 00:07:28,448 --> 00:07:33,801
  625. AI is a field of research
  626. to try to basically simulate
  627.  
  628. 136
  629. 00:07:33,844 --> 00:07:36,717
  630. all kinds of human capabilities.
  631.  
  632. 137
  633. 00:07:36,760 --> 00:07:38,719
  634. We're in the AI era.
  635.  
  636. 138
  637. 00:07:38,762 --> 00:07:40,503
  638. Silicon Valley
  639. has the ability to focus
  640.  
  641. 139
  642. 00:07:40,547 --> 00:07:42,462
  643. on one bright, shiny thing.
  644.  
  645. 140
  646. 00:07:42,505 --> 00:07:43,767
  647. It was social networking
  648.  
  649. 141
  650. 00:07:43,811 --> 00:07:45,508
  651. and social media
  652. over the last decade,
  653.  
  654. 142
  655. 00:07:45,552 --> 00:07:48,119
  656. and it's pretty clear
  657. that the bit has flipped.
  658.  
  659. 143
  660. 00:07:48,163 --> 00:07:50,557
  661. And it starts
  662. with machine learning.
  663.  
  664. 144
  665. 00:07:50,600 --> 00:07:54,343
  666. Nolan: When we look back at this
  667. moment, what was the first AI?
  668.  
  669. 145
  670. 00:07:54,386 --> 00:07:55,736
  671. It's not sexy,
  672. and it isn't the thing
  673.  
  674. 146
  675. 00:07:55,779 --> 00:07:57,389
  676. we could see at the movies,
  677.  
  678. 147
  679. 00:07:57,433 --> 00:08:00,741
  680. but you'd make a great case
  681. that Google created,
  682.  
  683. 148
  684. 00:08:00,784 --> 00:08:03,395
  685. not a search engine,
  686. but a godhead.
  687.  
  688. 149
  689. 00:08:03,439 --> 00:08:06,486
  690. A way for people to ask
  691. any question they wanted
  692.  
  693. 150
  694. 00:08:06,529 --> 00:08:08,270
  695. and get the answer they needed.
  696.  
  697. 151
  698. 00:08:08,313 --> 00:08:11,273
  699. Russell: Most people are not
  700. aware that what Google is doing
  701.  
  702. 152
  703. 00:08:11,316 --> 00:08:13,710
  704. is actually a form of
  705. artificial intelligence.
  706.  
  707. 153
  708. 00:08:13,754 --> 00:08:16,234
  709. They just go there,
  710. they type in a thing.
  711.  
  712. 154
  713. 00:08:16,278 --> 00:08:18,323
  714. Google gives them the answer.
  715.  
  716. 155
  717. 00:08:18,367 --> 00:08:21,544
  718. Musk: With each search,
  719. we train it to be better.
  720.  
  721. 156
  722. 00:08:21,588 --> 00:08:23,851
  723. Sometimes we're typing a search,
  724. and it tell us the answer
  725.  
  726. 157
  727. 00:08:23,894 --> 00:08:27,419
  728. before you've finished
  729. asking the question.
  730.  
  731. 158
  732. 00:08:27,463 --> 00:08:29,944
  733. You know, who is the president
  734. of Kazakhstan?
  735.  
  736. 159
  737. 00:08:29,987 --> 00:08:31,685
  738. And it'll just tell you.
  739.  
  740. 160
  741. 00:08:31,728 --> 00:08:33,600
  742. You don't have to go to the
  743. Kazakhstan national website
  744.  
  745. 161
  746. 00:08:33,643 --> 00:08:34,818
  747. to find out.
  748.  
  749. 162
  750. 00:08:34,862 --> 00:08:37,081
  751. You didn't used to be
  752. able to do that.
  753.  
  754. 163
  755. 00:08:37,125 --> 00:08:39,475
  756. Nolan:
  757. That is artificial intelligence.
  758.  
  759. 164
  760. 00:08:39,519 --> 00:08:42,783
  761. Years from now when we try
  762. to understand, we will say,
  763.  
  764. 165
  765. 00:08:42,826 --> 00:08:44,567
  766. "How did we miss it?"
  767.  
  768. 166
  769. 00:08:44,611 --> 00:08:47,527
  770. Markoff: It's one of
  771. the striking contradictions
  772.  
  773. 167
  774. 00:08:47,570 --> 00:08:48,484
  775. that we're facing.
  776.  
  777. 168
  778. 00:08:48,528 --> 00:08:50,051
  779. Google and Facebook, et al,
  780.  
  781. 169
  782. 00:08:50,094 --> 00:08:52,053
  783. have built businesses
  784. on giving us,
  785.  
  786. 170
  787. 00:08:52,096 --> 00:08:54,185
  788. as a society, free stuff.
  789.  
  790. 171
  791. 00:08:54,229 --> 00:08:56,013
  792. But it's a Faustian bargain.
  793.  
  794. 172
  795. 00:08:56,057 --> 00:09:00,017
  796. They're extracting something
  797. from us in exchange,
  798.  
  799. 173
  800. 00:09:00,061 --> 00:09:01,628
  801. but we don't know
  802.  
  803. 174
  804. 00:09:01,671 --> 00:09:03,760
  805. what code is running
  806. on the other side and why.
  807.  
  808. 175
  809. 00:09:03,804 --> 00:09:06,546
  810. We have no idea.
  811.  
  812. 176
  813. 00:09:06,589 --> 00:09:08,591
  814. It does strike
  815. right at the issue
  816.  
  817. 177
  818. 00:09:08,635 --> 00:09:11,028
  819. of how much we should
  820. trust these machines.
  821.  
  822. 178
  823. 00:09:14,162 --> 00:09:18,166
  824. I use computers
  825. literally for everything.
  826.  
  827. 179
  828. 00:09:18,209 --> 00:09:21,386
  829. There are so many
  830. computer advancements now,
  831.  
  832. 180
  833. 00:09:21,430 --> 00:09:23,824
  834. and it's become such
  835. a big part of our lives.
  836.  
  837. 181
  838. 00:09:23,867 --> 00:09:26,174
  839. It's just incredible
  840. what a computer can do.
  841.  
  842. 182
  843. 00:09:26,217 --> 00:09:29,090
  844. You can actually carry
  845. a computer in your purse.
  846.  
  847. 183
  848. 00:09:29,133 --> 00:09:31,571
  849. I mean, how awesome is that?
  850.  
  851. 184
  852. 00:09:31,614 --> 00:09:35,052
  853. I think most technology is meant
  854. to make things easier
  855.  
  856. 185
  857. 00:09:35,096 --> 00:09:37,315
  858. and simpler for all of us,
  859.  
  860. 186
  861. 00:09:37,359 --> 00:09:40,362
  862. so hopefully that just
  863. remains the focus.
  864.  
  865. 187
  866. 00:09:40,405 --> 00:09:43,147
  867. I think everybody loves
  868. their computers.
  869.  
  870. 188
  871. 00:09:44,409 --> 00:09:51,678
  872. ♪♪
  873.  
  874. 189
  875. 00:09:51,721 --> 00:09:53,810
  876. People don't realize
  877. they are constantly
  878.  
  879. 190
  880. 00:09:53,854 --> 00:09:59,076
  881. being negotiated with
  882. by machines,
  883.  
  884. 191
  885. 00:09:59,120 --> 00:10:02,993
  886. whether that's the price
  887. of products in your Amazon cart,
  888.  
  889. 192
  890. 00:10:03,037 --> 00:10:05,517
  891. whether you can get
  892. on a particular flight,
  893.  
  894. 193
  895. 00:10:05,561 --> 00:10:08,912
  896. whether you can reserve
  897. a room at a particular hotel.
  898.  
  899. 194
  900. 00:10:08,956 --> 00:10:11,959
  901. What you're experiencing
  902. are machine-learning algorithms
  903.  
  904. 195
  905. 00:10:12,002 --> 00:10:14,265
  906. that have determined
  907. that a person like you
  908.  
  909. 196
  910. 00:10:14,309 --> 00:10:15,919
  911. is willing to pay 2 cents more
  912.  
  913. 197
  914. 00:10:15,963 --> 00:10:17,791
  915. and is changing the price.
  916.  
  917. 198
  918. 00:10:17,834 --> 00:10:21,751
  919. ♪♪
  920.  
  921. 199
  922. 00:10:21,795 --> 00:10:24,014
  923. Kosinski: Now, a computer looks
  924. at millions of people
  925.  
  926. 200
  927. 00:10:24,058 --> 00:10:28,105
  928. simultaneously for
  929. very subtle patterns.
  930.  
  931. 201
  932. 00:10:28,149 --> 00:10:31,369
  933. You can take seemingly
  934. innocent digital footprints,
  935.  
  936. 202
  937. 00:10:31,413 --> 00:10:34,677
  938. such as someone's playlist
  939. on Spotify,
  940.  
  941. 203
  942. 00:10:34,721 --> 00:10:37,201
  943. or stuff that they
  944. bought on Amazon,
  945.  
  946. 204
  947. 00:10:37,245 --> 00:10:40,291
  948. and then use algorithms
  949. to translate this
  950.  
  951. 205
  952. 00:10:40,335 --> 00:10:44,513
  953. into a very detailed and a
  954. very accurate, intimate profile.
  955.  
  956. 206
  957. 00:10:47,603 --> 00:10:50,911
  958. Kaplan: There is a dossier on
  959. each of us that is so extensive
  960.  
  961. 207
  962. 00:10:50,954 --> 00:10:52,695
  963. it would be possibly
  964. accurate to say
  965.  
  966. 208
  967. 00:10:52,739 --> 00:10:55,698
  968. that they know more about you
  969. than your mother does.
  970.  
  971. 209
  972. 00:10:55,742 --> 00:11:04,054
  973. ♪♪
  974.  
  975. 210
  976. 00:11:04,098 --> 00:11:06,883
  977. Tegmark: The major cause
  978. of the recent AI breakthrough
  979.  
  980. 211
  981. 00:11:06,927 --> 00:11:08,580
  982. isn't just that some dude
  983.  
  984. 212
  985. 00:11:08,624 --> 00:11:11,583
  986. had a brilliant insight
  987. all of a sudden,
  988.  
  989. 213
  990. 00:11:11,627 --> 00:11:14,325
  991. but simply that we have
  992. much bigger data
  993.  
  994. 214
  995. 00:11:14,369 --> 00:11:18,242
  996. to train them on
  997. and vastly better computers.
  998.  
  999. 215
  1000. 00:11:18,286 --> 00:11:19,940
  1001. el Kaliouby:
  1002. The magic is in the data.
  1003.  
  1004. 216
  1005. 00:11:19,983 --> 00:11:21,463
  1006. It's a ton of data.
  1007.  
  1008. 217
  1009. 00:11:21,506 --> 00:11:23,726
  1010. I mean, it's data
  1011. that's never existed before.
  1012.  
  1013. 218
  1014. 00:11:23,770 --> 00:11:26,686
  1015. We've never had
  1016. this data before.
  1017.  
  1018. 219
  1019. 00:11:26,729 --> 00:11:30,733
  1020. We've created technologies
  1021. that allow us to capture
  1022.  
  1023. 220
  1024. 00:11:30,777 --> 00:11:33,040
  1025. vast amounts of information.
  1026.  
  1027. 221
  1028. 00:11:33,083 --> 00:11:35,738
  1029. If you think of a billion
  1030. cellphones on the planet
  1031.  
  1032. 222
  1033. 00:11:35,782 --> 00:11:38,393
  1034. with gyroscopes
  1035. and accelerometers
  1036.  
  1037. 223
  1038. 00:11:38,436 --> 00:11:39,786
  1039. and fingerprint readers...
  1040.  
  1041. 224
  1042. 00:11:39,829 --> 00:11:42,005
  1043. couple that with the GPS
  1044. and the photos they take
  1045.  
  1046. 225
  1047. 00:11:42,049 --> 00:11:43,964
  1048. and the tweets that you send,
  1049.  
  1050. 226
  1051. 00:11:44,007 --> 00:11:47,750
  1052. we're all giving off huge
  1053. amounts of data individually.
  1054.  
  1055. 227
  1056. 00:11:47,794 --> 00:11:50,274
  1057. Cars that drive as the cameras
  1058. on them suck up information
  1059.  
  1060. 228
  1061. 00:11:50,318 --> 00:11:52,059
  1062. about the world around them.
  1063.  
  1064. 229
  1065. 00:11:52,102 --> 00:11:54,844
  1066. The satellites that are now
  1067. in orbit the size of a toaster.
  1068.  
  1069. 230
  1070. 00:11:54,888 --> 00:11:57,629
  1071. The infrared about
  1072. the vegetation on the planet.
  1073.  
  1074. 231
  1075. 00:11:57,673 --> 00:11:59,109
  1076. The buoys that are out
  1077. in the oceans
  1078.  
  1079. 232
  1080. 00:11:59,153 --> 00:12:01,024
  1081. to feed into the climate models.
  1082.  
  1083. 233
  1084. 00:12:01,068 --> 00:12:05,028
  1085. ♪♪
  1086.  
  1087. 234
  1088. 00:12:05,072 --> 00:12:08,902
  1089. And the NSA, the CIA,
  1090. as they collect information
  1091.  
  1092. 235
  1093. 00:12:08,945 --> 00:12:12,644
  1094. about the
  1095. geopolitical situations.
  1096.  
  1097. 236
  1098. 00:12:12,688 --> 00:12:15,604
  1099. The world today is literally
  1100. swimming in this data.
  1101.  
  1102. 237
  1103. 00:12:15,647 --> 00:12:20,565
  1104. ♪♪
  1105.  
  1106. 238
  1107. 00:12:20,609 --> 00:12:22,480
  1108. Kosinski: Back in 2012,
  1109.  
  1110. 239
  1111. 00:12:22,524 --> 00:12:25,875
  1112. IBM estimated
  1113. that an average human being
  1114.  
  1115. 240
  1116. 00:12:25,919 --> 00:12:31,098
  1117. leaves 500 megabytes
  1118. of digital footprints every day.
  1119.  
  1120. 241
  1121. 00:12:31,141 --> 00:12:34,841
  1122. If you wanted to back up
  1123. on the one day worth of data
  1124.  
  1125. 242
  1126. 00:12:34,884 --> 00:12:36,494
  1127. that humanity produces
  1128.  
  1129. 243
  1130. 00:12:36,538 --> 00:12:39,062
  1131. and imprint it out
  1132. on a letter-sized paper,
  1133.  
  1134. 244
  1135. 00:12:39,106 --> 00:12:43,806
  1136. double-sided, font size 12,
  1137. and you stack it up,
  1138.  
  1139. 245
  1140. 00:12:43,850 --> 00:12:46,113
  1141. it would reach from
  1142. the surface of the Earth
  1143.  
  1144. 246
  1145. 00:12:46,156 --> 00:12:49,116
  1146. to the sun four times over.
  1147.  
  1148. 247
  1149. 00:12:49,159 --> 00:12:51,292
  1150. That's every day.
  1151.  
  1152. 248
  1153. 00:12:51,335 --> 00:12:53,816
  1154. Kaplan: The data itself
  1155. is not good or evil.
  1156.  
  1157. 249
  1158. 00:12:53,860 --> 00:12:55,470
  1159. It's how it's used.
  1160.  
  1161. 250
  1162. 00:12:55,513 --> 00:12:58,342
  1163. We're relying, really,
  1164. on the goodwill of these people
  1165.  
  1166. 251
  1167. 00:12:58,386 --> 00:13:01,171
  1168. and on the policies
  1169. of these companies.
  1170.  
  1171. 252
  1172. 00:13:01,215 --> 00:13:03,870
  1173. There is no legal requirement
  1174. for how they can
  1175.  
  1176. 253
  1177. 00:13:03,913 --> 00:13:06,307
  1178. and should use
  1179. that kind of data.
  1180.  
  1181. 254
  1182. 00:13:06,350 --> 00:13:09,266
  1183. That, to me, is at the heart
  1184. of the trust issue.
  1185.  
  1186. 255
  1187. 00:13:11,007 --> 00:13:13,793
  1188. Barrat: Right now there's a
  1189. giant race for creating machines
  1190.  
  1191. 256
  1192. 00:13:13,836 --> 00:13:15,751
  1193. that are as smart as humans.
  1194.  
  1195. 257
  1196. 00:13:15,795 --> 00:13:17,971
  1197. Google -- They're working on
  1198. what's really the kind of
  1199.  
  1200. 258
  1201. 00:13:18,014 --> 00:13:20,016
  1202. Manhattan Project
  1203. of artificial intelligence.
  1204.  
  1205. 259
  1206. 00:13:20,060 --> 00:13:22,671
  1207. They've got the most money.
  1208. They've got the most talent.
  1209.  
  1210. 260
  1211. 00:13:22,714 --> 00:13:27,067
  1212. They're buying up AI companies
  1213. and robotics companies.
  1214.  
  1215. 261
  1216. 00:13:27,110 --> 00:13:29,069
  1217. Urban: People still think
  1218. of Google as a search engine
  1219.  
  1220. 262
  1221. 00:13:29,112 --> 00:13:30,722
  1222. and their e-mail provider
  1223.  
  1224. 263
  1225. 00:13:30,766 --> 00:13:33,943
  1226. and a lot of other things
  1227. that we use on a daily basis,
  1228.  
  1229. 264
  1230. 00:13:33,987 --> 00:13:39,383
  1231. but behind that search box
  1232. are 10 million servers.
  1233.  
  1234. 265
  1235. 00:13:39,427 --> 00:13:42,299
  1236. That makes Google the most
  1237. powerful computing platform
  1238.  
  1239. 266
  1240. 00:13:42,343 --> 00:13:43,910
  1241. in the world.
  1242.  
  1243. 267
  1244. 00:13:43,953 --> 00:13:47,217
  1245. Google is now working
  1246. on an AI computing platform
  1247.  
  1248. 268
  1249. 00:13:47,261 --> 00:13:50,133
  1250. that will have
  1251. 100 million servers.
  1252.  
  1253. 269
  1254. 00:13:52,179 --> 00:13:53,963
  1255. So when you're interacting
  1256. with Google,
  1257.  
  1258. 270
  1259. 00:13:54,007 --> 00:13:56,052
  1260. we're just seeing
  1261. the toenail of something
  1262.  
  1263. 271
  1264. 00:13:56,096 --> 00:13:58,881
  1265. that is a giant beast
  1266. in the making.
  1267.  
  1268. 272
  1269. 00:13:58,925 --> 00:14:00,622
  1270. And the truth is,
  1271. I'm not even sure
  1272.  
  1273. 273
  1274. 00:14:00,665 --> 00:14:02,798
  1275. that Google knows
  1276. what it's becoming.
  1277.  
  1278. 274
  1279. 00:14:02,842 --> 00:14:11,502
  1280. ♪♪
  1281.  
  1282. 275
  1283. 00:14:11,546 --> 00:14:14,114
  1284. Phoenix: If you look inside of
  1285. what algorithms are being used
  1286.  
  1287. 276
  1288. 00:14:14,157 --> 00:14:15,811
  1289. at Google,
  1290.  
  1291. 277
  1292. 00:14:15,855 --> 00:14:20,076
  1293. it's technology
  1294. largely from the '80s.
  1295.  
  1296. 278
  1297. 00:14:20,120 --> 00:14:23,863
  1298. So these are models that you
  1299. train by showing them a 1, a 2,
  1300.  
  1301. 279
  1302. 00:14:23,906 --> 00:14:27,344
  1303. and a 3, and it learns not
  1304. what a 1 is or what a 2 is --
  1305.  
  1306. 280
  1307. 00:14:27,388 --> 00:14:30,434
  1308. It learns what the difference
  1309. between a 1 and a 2 is.
  1310.  
  1311. 281
  1312. 00:14:30,478 --> 00:14:32,436
  1313. It's just a computation.
  1314.  
  1315. 282
  1316. 00:14:32,480 --> 00:14:35,396
  1317. In the last half decade, where
  1318. we've made this rapid progress,
  1319.  
  1320. 283
  1321. 00:14:35,439 --> 00:14:38,268
  1322. it has all been
  1323. in pattern recognition.
  1324.  
  1325. 284
  1326. 00:14:38,312 --> 00:14:41,184
  1327. Tegmark: Most of
  1328. the good, old-fashioned AI
  1329.  
  1330. 285
  1331. 00:14:41,228 --> 00:14:44,057
  1332. was when we would tell
  1333. our computers
  1334.  
  1335. 286
  1336. 00:14:44,100 --> 00:14:46,798
  1337. how to play a game like chess...
  1338.  
  1339. 287
  1340. 00:14:46,842 --> 00:14:49,584
  1341. from the old paradigm where
  1342. you just tell the computer
  1343.  
  1344. 288
  1345. 00:14:49,627 --> 00:14:52,195
  1346. exactly what to do.
  1347.  
  1348. 289
  1349. 00:14:54,502 --> 00:14:57,505
  1350. Announcer:
  1351. This is "Jeopardy!"
  1352.  
  1353. 290
  1354. 00:14:57,548 --> 00:14:59,376
  1355. ♪♪
  1356.  
  1357. 291
  1358. 00:14:59,420 --> 00:15:02,510
  1359. "The IBM Challenge"!
  1360.  
  1361. 292
  1362. 00:15:02,553 --> 00:15:05,730
  1363. Ferrucci: No one at the time
  1364. had thought that a machine
  1365.  
  1366. 293
  1367. 00:15:05,774 --> 00:15:08,298
  1368. could have the precision
  1369. and the confidence
  1370.  
  1371. 294
  1372. 00:15:08,342 --> 00:15:09,952
  1373. and the speed
  1374. to play "Jeopardy!"
  1375.  
  1376. 295
  1377. 00:15:09,996 --> 00:15:11,475
  1378. well enough against
  1379. the best humans.
  1380.  
  1381. 296
  1382. 00:15:11,519 --> 00:15:14,609
  1383. Let's play "Jeopardy!"
  1384.  
  1385. 297
  1386. 00:15:18,569 --> 00:15:20,354
  1387. Watson.Watson: What is "shoe"?
  1388.  
  1389. 298
  1390. 00:15:20,397 --> 00:15:21,877
  1391. You are right.
  1392. You get to pick.
  1393.  
  1394. 299
  1395. 00:15:21,921 --> 00:15:24,836
  1396. Literary Character APB
  1397. for $800.
  1398.  
  1399. 300
  1400. 00:15:24,880 --> 00:15:28,014
  1401. Answer --
  1402. the Daily Double.
  1403.  
  1404. 301
  1405. 00:15:28,057 --> 00:15:31,539
  1406. Watson actually got its
  1407. knowledge by reading Wikipedia
  1408.  
  1409. 302
  1410. 00:15:31,582 --> 00:15:34,672
  1411. and 200 million pages
  1412. of natural-language documents.
  1413.  
  1414. 303
  1415. 00:15:34,716 --> 00:15:36,674
  1416. Ferrucci:
  1417. You can't program every line
  1418.  
  1419. 304
  1420. 00:15:36,718 --> 00:15:38,502
  1421. of how the world works.
  1422.  
  1423. 305
  1424. 00:15:38,546 --> 00:15:40,722
  1425. The machine has to learn
  1426. by reading.
  1427.  
  1428. 306
  1429. 00:15:40,765 --> 00:15:42,202
  1430. Now we come to Watson.
  1431.  
  1432. 307
  1433. 00:15:42,245 --> 00:15:43,986
  1434. "Who is Bram Stoker?"
  1435.  
  1436. 308
  1437. 00:15:44,030 --> 00:15:45,988
  1438. And the wager?
  1439.  
  1440. 309
  1441. 00:15:46,032 --> 00:15:49,165
  1442. Hello! $17,973.
  1443.  
  1444. 310
  1445. 00:15:49,209 --> 00:15:50,993
  1446. $41,413.
  1447.  
  1448. 311
  1449. 00:15:51,037 --> 00:15:53,343
  1450. And a two-day total
  1451. of $77--
  1452.  
  1453. 312
  1454. 00:15:53,387 --> 00:15:56,694
  1455. Phoenix: Watson's trained
  1456. on huge amounts of text,
  1457.  
  1458. 313
  1459. 00:15:56,738 --> 00:15:59,828
  1460. but it's not like it
  1461. understands what it's saying.
  1462.  
  1463. 314
  1464. 00:15:59,871 --> 00:16:02,309
  1465. It doesn't know that water makes
  1466. things wet by touching water
  1467.  
  1468. 315
  1469. 00:16:02,352 --> 00:16:04,441
  1470. and by seeing the way
  1471. things behave in the world
  1472.  
  1473. 316
  1474. 00:16:04,485 --> 00:16:06,182
  1475. the way you and I do.
  1476.  
  1477. 317
  1478. 00:16:06,226 --> 00:16:10,143
  1479. A lot of language AI today
  1480. is not building logical models
  1481.  
  1482. 318
  1483. 00:16:10,186 --> 00:16:11,622
  1484. of how the world works.
  1485.  
  1486. 319
  1487. 00:16:11,666 --> 00:16:15,365
  1488. Rather, it's looking at
  1489. how the words appear
  1490.  
  1491. 320
  1492. 00:16:15,409 --> 00:16:18,238
  1493. in the context of other words.
  1494.  
  1495. 321
  1496. 00:16:18,281 --> 00:16:20,196
  1497. Barrat: David Ferrucci
  1498. developed IBM's Watson,
  1499.  
  1500. 322
  1501. 00:16:20,240 --> 00:16:23,547
  1502. and somebody asked him,
  1503. "Does Watson think?"
  1504.  
  1505. 323
  1506. 00:16:23,591 --> 00:16:27,160
  1507. And he said,
  1508. "Does a submarine swim?"
  1509.  
  1510. 324
  1511. 00:16:27,203 --> 00:16:29,031
  1512. And what they meant was,
  1513. when they developed submarines,
  1514.  
  1515. 325
  1516. 00:16:29,075 --> 00:16:32,992
  1517. they borrowed basic principles
  1518. of swimming from fish.
  1519.  
  1520. 326
  1521. 00:16:33,035 --> 00:16:35,037
  1522. But a submarine swims
  1523. farther and faster than fish
  1524.  
  1525. 327
  1526. 00:16:35,081 --> 00:16:36,125
  1527. and can carry a huge payload.
  1528.  
  1529. 328
  1530. 00:16:36,169 --> 00:16:39,911
  1531. It out-swims fish.
  1532.  
  1533. 329
  1534. 00:16:39,955 --> 00:16:41,870
  1535. Ng: Watson winning the game
  1536. of "Jeopardy!"
  1537.  
  1538. 330
  1539. 00:16:41,913 --> 00:16:43,741
  1540. will go down
  1541. in the history of AI
  1542.  
  1543. 331
  1544. 00:16:43,785 --> 00:16:46,570
  1545. as a significant milestone.
  1546.  
  1547. 332
  1548. 00:16:46,614 --> 00:16:49,269
  1549. We tend to be amazed
  1550. when the machine does so well.
  1551.  
  1552. 333
  1553. 00:16:49,312 --> 00:16:52,663
  1554. I'm even more amazed when the
  1555. computer beats humans at things
  1556.  
  1557. 334
  1558. 00:16:52,707 --> 00:16:55,188
  1559. that humans are
  1560. naturally good at.
  1561.  
  1562. 335
  1563. 00:16:55,231 --> 00:16:58,060
  1564. This is how we make progress.
  1565.  
  1566. 336
  1567. 00:16:58,104 --> 00:17:00,671
  1568. In the early days of
  1569. the Google Brain project,
  1570.  
  1571. 337
  1572. 00:17:00,715 --> 00:17:02,804
  1573. I gave the team a very
  1574. simple instruction,
  1575.  
  1576. 338
  1577. 00:17:02,847 --> 00:17:05,807
  1578. which was, "Build the biggest
  1579. neural network possible,
  1580.  
  1581. 339
  1582. 00:17:05,850 --> 00:17:08,157
  1583. like 1,000 computers."
  1584.  
  1585. 340
  1586. 00:17:08,201 --> 00:17:09,724
  1587. Musk: A neural net is
  1588. something very close
  1589.  
  1590. 341
  1591. 00:17:09,767 --> 00:17:12,161
  1592. to a simulation
  1593. of how the brain works.
  1594.  
  1595. 342
  1596. 00:17:12,205 --> 00:17:16,818
  1597. It's very probabilistic,
  1598. but with contextual relevance.
  1599.  
  1600. 343
  1601. 00:17:16,861 --> 00:17:18,298
  1602. Urban: In your brain,
  1603. you have long neurons
  1604.  
  1605. 344
  1606. 00:17:18,341 --> 00:17:20,256
  1607. that connect to thousands
  1608. of other neurons,
  1609.  
  1610. 345
  1611. 00:17:20,300 --> 00:17:22,519
  1612. and you have these pathways
  1613. that are formed and forged
  1614.  
  1615. 346
  1616. 00:17:22,563 --> 00:17:24,739
  1617. based on what
  1618. the brain needs to do.
  1619.  
  1620. 347
  1621. 00:17:24,782 --> 00:17:28,960
  1622. When a baby tries something and
  1623. it succeeds, there's a reward,
  1624.  
  1625. 348
  1626. 00:17:29,004 --> 00:17:32,312
  1627. and that pathway that created
  1628. the success is strengthened.
  1629.  
  1630. 349
  1631. 00:17:32,355 --> 00:17:34,662
  1632. If it fails at something,
  1633. the pathway is weakened,
  1634.  
  1635. 350
  1636. 00:17:34,705 --> 00:17:36,794
  1637. and so, over time,
  1638. the brain becomes honed
  1639.  
  1640. 351
  1641. 00:17:36,838 --> 00:17:40,320
  1642. to be good at
  1643. the environment around it.
  1644.  
  1645. 352
  1646. 00:17:40,363 --> 00:17:43,279
  1647. Ng: Really, it's just getting
  1648. machines to learn by themselves.
  1649.  
  1650. 353
  1651. 00:17:43,323 --> 00:17:45,238
  1652. This is called "deep learning,"
  1653. and "deep learning"
  1654.  
  1655. 354
  1656. 00:17:45,281 --> 00:17:48,676
  1657. and "neural networks"
  1658. mean roughly the same thing.
  1659.  
  1660. 355
  1661. 00:17:48,719 --> 00:17:52,375
  1662. Tegmark: Deep learning
  1663. is a totally different approach
  1664.  
  1665. 356
  1666. 00:17:52,419 --> 00:17:55,161
  1667. where the computer learns
  1668. more like a toddler,
  1669.  
  1670. 357
  1671. 00:17:55,204 --> 00:17:56,466
  1672. by just getting a lot of data
  1673.  
  1674. 358
  1675. 00:17:56,510 --> 00:18:00,340
  1676. and eventually
  1677. figuring stuff out.
  1678.  
  1679. 359
  1680. 00:18:00,383 --> 00:18:03,125
  1681. The computer just gets
  1682. smarter and smarter
  1683.  
  1684. 360
  1685. 00:18:03,169 --> 00:18:05,997
  1686. as it has more experiences.
  1687.  
  1688. 361
  1689. 00:18:06,041 --> 00:18:08,130
  1690. Ng: So, imagine, if you will,
  1691. a neural network, you know,
  1692.  
  1693. 362
  1694. 00:18:08,174 --> 00:18:09,697
  1695. like 1,000 computers.
  1696.  
  1697. 363
  1698. 00:18:09,740 --> 00:18:11,438
  1699. And it wakes up
  1700. not knowing anything.
  1701.  
  1702. 364
  1703. 00:18:11,481 --> 00:18:14,093
  1704. And we made it watch YouTube
  1705. for a week.
  1706.  
  1707. 365
  1708. 00:18:14,136 --> 00:18:16,704
  1709. ♪♪
  1710.  
  1711. 366
  1712. 00:18:18,706 --> 00:18:20,360
  1713. ♪ Oppan Gangnam style
  1714.  
  1715. 367
  1716. 00:18:20,403 --> 00:18:23,189
  1717. Ow!
  1718.  
  1719. 368
  1720. 00:18:25,408 --> 00:18:28,194
  1721. Charlie!
  1722. That really hurt!
  1723.  
  1724. 369
  1725. 00:18:28,237 --> 00:18:30,152
  1726. ♪♪
  1727.  
  1728. 370
  1729. 00:18:30,196 --> 00:18:31,327
  1730. ♪ Gangnam style
  1731.  
  1732. 371
  1733. 00:18:31,371 --> 00:18:33,286
  1734. ♪ Op, op, op, op
  1735.  
  1736. 372
  1737. 00:18:33,329 --> 00:18:36,202
  1738. ♪ Oppan Gangnam style
  1739.  
  1740. 373
  1741. 00:18:36,245 --> 00:18:38,508
  1742. Ng: And so, after watching
  1743. YouTube for a week,
  1744.  
  1745. 374
  1746. 00:18:38,552 --> 00:18:39,988
  1747. what would it learn?
  1748.  
  1749. 375
  1750. 00:18:40,031 --> 00:18:41,903
  1751. We had a hypothesis that
  1752. it would learn to detect
  1753.  
  1754. 376
  1755. 00:18:41,946 --> 00:18:44,384
  1756. commonly occurring objects
  1757. in videos.
  1758.  
  1759. 377
  1760. 00:18:44,427 --> 00:18:47,517
  1761. And so, we know that human faces
  1762. appear a lot in videos,
  1763.  
  1764. 378
  1765. 00:18:47,561 --> 00:18:49,302
  1766. so we looked,
  1767. and, lo and behold,
  1768.  
  1769. 379
  1770. 00:18:49,345 --> 00:18:51,608
  1771. there was a neuron that had
  1772. learned to detect human faces.
  1773.  
  1774. 380
  1775. 00:18:51,652 --> 00:18:56,265
  1776. Leave Britney alone!
  1777.  
  1778. 381
  1779. 00:18:56,309 --> 00:18:58,354
  1780. Well, what else
  1781. appears in videos a lot?
  1782.  
  1783. 382
  1784. 00:19:00,095 --> 00:19:01,792
  1785. So, we looked,
  1786. and to our surprise,
  1787.  
  1788. 383
  1789. 00:19:01,836 --> 00:19:04,882
  1790. there was actually a neuron
  1791. that had learned to detect cats.
  1792.  
  1793. 384
  1794. 00:19:04,926 --> 00:19:14,849
  1795. ♪♪
  1796.  
  1797. 385
  1798. 00:19:14,892 --> 00:19:17,068
  1799. I still remember
  1800. seeing recognition.
  1801.  
  1802. 386
  1803. 00:19:17,112 --> 00:19:18,635
  1804. "Wow, that's a cat.
  1805. Okay, cool.
  1806.  
  1807. 387
  1808. 00:19:18,679 --> 00:19:20,071
  1809. Great."
  1810.  
  1811. 388
  1812. 00:19:23,162 --> 00:19:24,859
  1813. Barrat:
  1814. It's all pretty innocuous
  1815.  
  1816. 389
  1817. 00:19:24,902 --> 00:19:26,295
  1818. when you're thinking
  1819. about the future.
  1820.  
  1821. 390
  1822. 00:19:26,339 --> 00:19:29,733
  1823. It all seems kind of
  1824. harmless and benign.
  1825.  
  1826. 391
  1827. 00:19:29,777 --> 00:19:31,605
  1828. But we're making
  1829. cognitive architectures
  1830.  
  1831. 392
  1832. 00:19:31,648 --> 00:19:33,520
  1833. that will fly farther
  1834. and faster than us
  1835.  
  1836. 393
  1837. 00:19:33,563 --> 00:19:35,086
  1838. and carry a bigger payload,
  1839.  
  1840. 394
  1841. 00:19:35,130 --> 00:19:37,437
  1842. and they won't be
  1843. warm and fuzzy.
  1844.  
  1845. 395
  1846. 00:19:37,480 --> 00:19:39,656
  1847. Ferrucci: I think that,
  1848. in three to five years,
  1849.  
  1850. 396
  1851. 00:19:39,700 --> 00:19:41,702
  1852. you will see a computer system
  1853.  
  1854. 397
  1855. 00:19:41,745 --> 00:19:45,401
  1856. that will be able
  1857. to autonomously learn
  1858.  
  1859. 398
  1860. 00:19:45,445 --> 00:19:49,013
  1861. how to understand,
  1862. how to build understanding,
  1863.  
  1864. 399
  1865. 00:19:49,057 --> 00:19:51,364
  1866. not unlike the way
  1867. the human mind works.
  1868.  
  1869. 400
  1870. 00:19:53,931 --> 00:19:56,891
  1871. Whatever that lunch was,
  1872. it was certainly delicious.
  1873.  
  1874. 401
  1875. 00:19:56,934 --> 00:19:59,807
  1876. Simply some of
  1877. Robby's synthetics.
  1878.  
  1879. 402
  1880. 00:19:59,850 --> 00:20:01,635
  1881. He's your cook, too?
  1882.  
  1883. 403
  1884. 00:20:01,678 --> 00:20:04,551
  1885. Even manufactures
  1886. the raw materials.
  1887.  
  1888. 404
  1889. 00:20:04,594 --> 00:20:06,944
  1890. Come around here, Robby.
  1891.  
  1892. 405
  1893. 00:20:06,988 --> 00:20:09,773
  1894. I'll show you
  1895. how this works.
  1896.  
  1897. 406
  1898. 00:20:11,122 --> 00:20:13,342
  1899. One introduces
  1900. a sample of human food
  1901.  
  1902. 407
  1903. 00:20:13,386 --> 00:20:15,344
  1904. through this aperture.
  1905.  
  1906. 408
  1907. 00:20:15,388 --> 00:20:17,738
  1908. Down here there's a small
  1909. built-in chemical laboratory,
  1910.  
  1911. 409
  1912. 00:20:17,781 --> 00:20:19,218
  1913. where he analyzes it.
  1914.  
  1915. 410
  1916. 00:20:19,261 --> 00:20:21,263
  1917. Later, he can reproduce
  1918. identical molecules
  1919.  
  1920. 411
  1921. 00:20:21,307 --> 00:20:22,482
  1922. in any shape or quantity.
  1923.  
  1924. 412
  1925. 00:20:22,525 --> 00:20:24,614
  1926. Why, it's
  1927. a housewife's dream.
  1928.  
  1929. 413
  1930. 00:20:24,658 --> 00:20:26,834
  1931. Announcer: Meet Baxter,
  1932.  
  1933. 414
  1934. 00:20:26,877 --> 00:20:29,445
  1935. revolutionary
  1936. new category of robots,
  1937.  
  1938. 415
  1939. 00:20:29,489 --> 00:20:30,490
  1940. with common sense.
  1941.  
  1942. 416
  1943. 00:20:30,533 --> 00:20:31,839
  1944. Baxter...
  1945.  
  1946. 417
  1947. 00:20:31,882 --> 00:20:33,449
  1948. Barrat: Baxter is
  1949. a really good example
  1950.  
  1951. 418
  1952. 00:20:33,493 --> 00:20:36,887
  1953. of the kind of competition
  1954. we face from machines.
  1955.  
  1956. 419
  1957. 00:20:36,931 --> 00:20:42,676
  1958. Baxter can do almost anything
  1959. we can do with our hands.
  1960.  
  1961. 420
  1962. 00:20:42,719 --> 00:20:45,722
  1963. Baxter costs about
  1964. what a minimum-wage worker
  1965.  
  1966. 421
  1967. 00:20:45,766 --> 00:20:47,507
  1968. makes in a year.
  1969.  
  1970. 422
  1971. 00:20:47,550 --> 00:20:48,769
  1972. But Baxter won't be
  1973. taking the place
  1974.  
  1975. 423
  1976. 00:20:48,812 --> 00:20:50,118
  1977. of one minimum-wage worker --
  1978.  
  1979. 424
  1980. 00:20:50,161 --> 00:20:51,772
  1981. He'll be taking
  1982. the place of three,
  1983.  
  1984. 425
  1985. 00:20:51,815 --> 00:20:55,515
  1986. because they never get tired,
  1987. they never take breaks.
  1988.  
  1989. 426
  1990. 00:20:55,558 --> 00:20:57,865
  1991. Gourley: That's probably the
  1992. first thing we're gonna see --
  1993.  
  1994. 427
  1995. 00:20:57,908 --> 00:20:59,475
  1996. displacement of jobs.
  1997.  
  1998. 428
  1999. 00:20:59,519 --> 00:21:01,651
  2000. They're gonna be done
  2001. quicker, faster, cheaper
  2002.  
  2003. 429
  2004. 00:21:01,695 --> 00:21:04,088
  2005. by machines.
  2006.  
  2007. 430
  2008. 00:21:04,132 --> 00:21:07,657
  2009. Our ability to even stay current
  2010. is so insanely limited
  2011.  
  2012. 431
  2013. 00:21:07,701 --> 00:21:10,138
  2014. compared to
  2015. the machines we build.
  2016.  
  2017. 432
  2018. 00:21:10,181 --> 00:21:13,446
  2019. For example, now we have this
  2020. great movement of Uber and Lyft
  2021.  
  2022. 433
  2023. 00:21:13,489 --> 00:21:15,056
  2024. kind of making
  2025. transportation cheaper
  2026.  
  2027. 434
  2028. 00:21:15,099 --> 00:21:16,405
  2029. and democratizing
  2030. transportation,
  2031.  
  2032. 435
  2033. 00:21:16,449 --> 00:21:17,711
  2034. which is great.
  2035.  
  2036. 436
  2037. 00:21:17,754 --> 00:21:19,321
  2038. The next step is gonna be
  2039.  
  2040. 437
  2041. 00:21:19,365 --> 00:21:21,149
  2042. that they're all gonna be
  2043. replaced by driverless cars,
  2044.  
  2045. 438
  2046. 00:21:21,192 --> 00:21:22,411
  2047. and then all the Uber
  2048. and Lyft drivers
  2049.  
  2050. 439
  2051. 00:21:22,455 --> 00:21:25,936
  2052. have to find
  2053. something new to do.
  2054.  
  2055. 440
  2056. 00:21:25,980 --> 00:21:28,156
  2057. Barrat: There are
  2058. 4 million professional drivers
  2059.  
  2060. 441
  2061. 00:21:28,199 --> 00:21:29,723
  2062. in the United States.
  2063.  
  2064. 442
  2065. 00:21:29,766 --> 00:21:31,638
  2066. They're unemployed soon.
  2067.  
  2068. 443
  2069. 00:21:31,681 --> 00:21:34,075
  2070. 7 million people
  2071. that do data entry.
  2072.  
  2073. 444
  2074. 00:21:34,118 --> 00:21:37,339
  2075. Those people
  2076. are gonna be jobless.
  2077.  
  2078. 445
  2079. 00:21:37,383 --> 00:21:40,342
  2080. A job isn't just about money,
  2081. right?
  2082.  
  2083. 446
  2084. 00:21:40,386 --> 00:21:42,605
  2085. On a biological level,
  2086. it serves a purpose.
  2087.  
  2088. 447
  2089. 00:21:42,649 --> 00:21:45,391
  2090. It becomes a defining thing.
  2091.  
  2092. 448
  2093. 00:21:45,434 --> 00:21:48,350
  2094. When the jobs went away
  2095. in any given civilization,
  2096.  
  2097. 449
  2098. 00:21:48,394 --> 00:21:50,787
  2099. it doesn't take long
  2100. until that turns into violence.
  2101.  
  2102. 450
  2103. 00:21:53,355 --> 00:21:57,011
  2104. ♪♪
  2105.  
  2106. 451
  2107. 00:21:59,622 --> 00:22:02,016
  2108. We face a giant divide
  2109. between rich and poor,
  2110.  
  2111. 452
  2112. 00:22:02,059 --> 00:22:05,019
  2113. because that's what automation
  2114. and AI will provoke --
  2115.  
  2116. 453
  2117. 00:22:05,062 --> 00:22:08,588
  2118. a greater divide between
  2119. the haves and the have-nots.
  2120.  
  2121. 454
  2122. 00:22:08,631 --> 00:22:10,807
  2123. Right now, it's working
  2124. into the middle class,
  2125.  
  2126. 455
  2127. 00:22:10,851 --> 00:22:12,896
  2128. into white-collar jobs.
  2129.  
  2130. 456
  2131. 00:22:12,940 --> 00:22:15,334
  2132. IBM's Watson does
  2133. business analytics
  2134.  
  2135. 457
  2136. 00:22:15,377 --> 00:22:20,600
  2137. that we used to pay a business
  2138. analyst $300 an hour to do.
  2139.  
  2140. 458
  2141. 00:22:20,643 --> 00:22:23,037
  2142. Gourley: Today, you're going
  2143. to college to be a doctor,
  2144.  
  2145. 459
  2146. 00:22:23,080 --> 00:22:25,082
  2147. to be an accountant,
  2148. to be a journalist.
  2149.  
  2150. 460
  2151. 00:22:25,126 --> 00:22:28,608
  2152. It's unclear that there's
  2153. gonna be jobs there for you.
  2154.  
  2155. 461
  2156. 00:22:28,651 --> 00:22:32,612
  2157. Ng: If someone's planning for
  2158. a 40-year career in radiology,
  2159.  
  2160. 462
  2161. 00:22:32,655 --> 00:22:34,222
  2162. just reading images,
  2163.  
  2164. 463
  2165. 00:22:34,265 --> 00:22:35,745
  2166. I think that could be
  2167. a challenge
  2168.  
  2169. 464
  2170. 00:22:35,789 --> 00:22:36,920
  2171. to the new graduates of today.
  2172.  
  2173. 465
  2174. 00:22:39,270 --> 00:22:49,193
  2175. ♪♪
  2176.  
  2177. 466
  2178. 00:22:50,847 --> 00:22:58,464
  2179. ♪♪
  2180.  
  2181. 467
  2182. 00:22:58,507 --> 00:23:02,729
  2183. Dr. Herman: The da Vinci robot
  2184. is currently utilized
  2185.  
  2186. 468
  2187. 00:23:02,772 --> 00:23:07,516
  2188. by a variety of surgeons
  2189. for its accuracy and its ability
  2190.  
  2191. 469
  2192. 00:23:07,560 --> 00:23:12,303
  2193. to avoid the inevitable
  2194. fluctuations of the human hand.
  2195.  
  2196. 470
  2197. 00:23:12,347 --> 00:23:17,787
  2198. ♪♪
  2199.  
  2200. 471
  2201. 00:23:17,831 --> 00:23:23,358
  2202. ♪♪
  2203.  
  2204. 472
  2205. 00:23:23,402 --> 00:23:28,494
  2206. Anybody who watches this
  2207. feels the amazingness of it.
  2208.  
  2209. 473
  2210. 00:23:30,931 --> 00:23:34,674
  2211. You look through the scope,
  2212. and you're seeing the claw hand
  2213.  
  2214. 474
  2215. 00:23:34,717 --> 00:23:36,893
  2216. holding that woman's ovary.
  2217.  
  2218. 475
  2219. 00:23:36,937 --> 00:23:42,638
  2220. Humanity was resting right here
  2221. in the hands of this robot.
  2222.  
  2223. 476
  2224. 00:23:42,682 --> 00:23:46,947
  2225. People say it's the future,
  2226. but it's not the future --
  2227.  
  2228. 477
  2229. 00:23:46,990 --> 00:23:50,516
  2230. It's the present.
  2231.  
  2232. 478
  2233. 00:23:50,559 --> 00:23:52,474
  2234. Zilis: If you think about
  2235. a surgical robot,
  2236.  
  2237. 479
  2238. 00:23:52,518 --> 00:23:54,737
  2239. there's often not a lot
  2240. of intelligence in these things,
  2241.  
  2242. 480
  2243. 00:23:54,781 --> 00:23:56,783
  2244. but over time, as we put
  2245. more and more intelligence
  2246.  
  2247. 481
  2248. 00:23:56,826 --> 00:23:58,567
  2249. into these systems,
  2250.  
  2251. 482
  2252. 00:23:58,611 --> 00:24:02,441
  2253. the surgical robots can actually
  2254. learn from each robot surgery.
  2255.  
  2256. 483
  2257. 00:24:02,484 --> 00:24:04,181
  2258. They're tracking the movements,
  2259. they're understanding
  2260.  
  2261. 484
  2262. 00:24:04,225 --> 00:24:05,966
  2263. what worked
  2264. and what didn't work.
  2265.  
  2266. 485
  2267. 00:24:06,009 --> 00:24:08,708
  2268. And eventually, the robot
  2269. for routine surgeries
  2270.  
  2271. 486
  2272. 00:24:08,751 --> 00:24:12,320
  2273. is going to be able to perform
  2274. that entirely by itself...
  2275.  
  2276. 487
  2277. 00:24:12,363 --> 00:24:13,756
  2278. or with human supervision.
  2279.  
  2280. 488
  2281. 00:24:32,558 --> 00:24:34,995
  2282. ♪♪
  2283.  
  2284. 489
  2285. 00:24:35,038 --> 00:24:37,214
  2286. Dr. Herman: It seems that we're
  2287. feeding it and creating it,
  2288.  
  2289. 490
  2290. 00:24:37,258 --> 00:24:42,785
  2291. but, in a way, we are a slave
  2292. to the technology,
  2293.  
  2294. 491
  2295. 00:24:42,829 --> 00:24:45,701
  2296. because we can't go back.
  2297.  
  2298. 492
  2299. 00:24:50,053 --> 00:24:52,882
  2300. Gourley: The machines are taking
  2301. bigger and bigger bites
  2302.  
  2303. 493
  2304. 00:24:52,926 --> 00:24:57,147
  2305. out of our skill set
  2306. at an ever-increasing speed.
  2307.  
  2308. 494
  2309. 00:24:57,191 --> 00:24:59,236
  2310. And so we've got to run
  2311. faster and faster
  2312.  
  2313. 495
  2314. 00:24:59,280 --> 00:25:00,890
  2315. to keep ahead of the machines.
  2316.  
  2317. 496
  2318. 00:25:02,675 --> 00:25:04,677
  2319. How do I look?
  2320.  
  2321. 497
  2322. 00:25:04,720 --> 00:25:06,374
  2323. Good.
  2324.  
  2325. 498
  2326. 00:25:10,030 --> 00:25:11,553
  2327. Are you attracted to me?
  2328.  
  2329. 499
  2330. 00:25:11,597 --> 00:25:14,251
  2331. What?Are you attracted to me?
  2332.  
  2333. 500
  2334. 00:25:14,295 --> 00:25:17,777
  2335. You give me indications
  2336. that you are.
  2337.  
  2338. 501
  2339. 00:25:17,820 --> 00:25:20,562
  2340. I do?
  2341. Yes.
  2342.  
  2343. 502
  2344. 00:25:20,606 --> 00:25:22,608
  2345. Nolan: This is the future
  2346. we're headed into.
  2347.  
  2348. 503
  2349. 00:25:22,651 --> 00:25:26,046
  2350. We want to design
  2351. our companions.
  2352.  
  2353. 504
  2354. 00:25:26,089 --> 00:25:29,266
  2355. We're gonna like to see
  2356. a human face on AI.
  2357.  
  2358. 505
  2359. 00:25:29,310 --> 00:25:33,967
  2360. Therefore, gaming our emotions
  2361. will be depressingly easy.
  2362.  
  2363. 506
  2364. 00:25:34,010 --> 00:25:35,272
  2365. We're not that complicated.
  2366.  
  2367. 507
  2368. 00:25:35,316 --> 00:25:38,101
  2369. We're simple.
  2370. Stimulus-response.
  2371.  
  2372. 508
  2373. 00:25:38,145 --> 00:25:43,063
  2374. I can make you like me basically
  2375. by smiling at you a lot.
  2376.  
  2377. 509
  2378. 00:25:43,106 --> 00:25:45,674
  2379. AIs are gonna be fantastic
  2380. at manipulating us.
  2381.  
  2382. 510
  2383. 00:25:45,718 --> 00:25:54,640
  2384. ♪♪
  2385.  
  2386. 511
  2387. 00:25:54,683 --> 00:25:56,946
  2388. So, you've developed
  2389. a technology
  2390.  
  2391. 512
  2392. 00:25:56,990 --> 00:26:00,036
  2393. that can sense
  2394. what people are feeling.
  2395.  
  2396. 513
  2397. 00:26:00,080 --> 00:26:01,472
  2398. Right.
  2399. We've developed technology
  2400.  
  2401. 514
  2402. 00:26:01,516 --> 00:26:03,387
  2403. that can read
  2404. your facial expressions
  2405.  
  2406. 515
  2407. 00:26:03,431 --> 00:26:06,521
  2408. and map that to a number
  2409. of emotional states.
  2410.  
  2411. 516
  2412. 00:26:06,565 --> 00:26:08,697
  2413. el Kaliouby: 15 years ago,
  2414. I had just finished
  2415.  
  2416. 517
  2417. 00:26:08,741 --> 00:26:11,482
  2418. my undergraduate studies
  2419. in computer science,
  2420.  
  2421. 518
  2422. 00:26:11,526 --> 00:26:15,008
  2423. and it struck me that I was
  2424. spending a lot of time
  2425.  
  2426. 519
  2427. 00:26:15,051 --> 00:26:17,793
  2428. interacting with my laptops
  2429. and my devices,
  2430.  
  2431. 520
  2432. 00:26:17,837 --> 00:26:23,582
  2433. yet these devices had absolutely
  2434. no clue how I was feeling.
  2435.  
  2436. 521
  2437. 00:26:23,625 --> 00:26:26,802
  2438. I started thinking, "What if
  2439. this device could sense
  2440.  
  2441. 522
  2442. 00:26:26,846 --> 00:26:29,326
  2443. that I was stressed
  2444. or I was having a bad day?
  2445.  
  2446. 523
  2447. 00:26:29,370 --> 00:26:31,067
  2448. What would that open up?"
  2449.  
  2450. 524
  2451. 00:26:32,721 --> 00:26:34,418
  2452. Hi, first-graders!
  2453.  
  2454. 525
  2455. 00:26:34,462 --> 00:26:35,855
  2456. How are you?
  2457.  
  2458. 526
  2459. 00:26:35,898 --> 00:26:37,813
  2460. Can I get a hug?
  2461.  
  2462. 527
  2463. 00:26:37,857 --> 00:26:40,773
  2464. We had kids interact
  2465. with the technology.
  2466.  
  2467. 528
  2468. 00:26:40,816 --> 00:26:42,862
  2469. A lot of it
  2470. is still in development,
  2471.  
  2472. 529
  2473. 00:26:42,905 --> 00:26:44,472
  2474. but it was just amazing.
  2475.  
  2476. 530
  2477. 00:26:44,515 --> 00:26:46,648
  2478. Who likes robots?
  2479. Me!
  2480.  
  2481. 531
  2482. 00:26:46,692 --> 00:26:48,911
  2483. Who wants to have a robot
  2484. in their house?
  2485.  
  2486. 532
  2487. 00:26:48,955 --> 00:26:51,479
  2488. What would you use
  2489. a robot for, Jack?
  2490.  
  2491. 533
  2492. 00:26:51,522 --> 00:26:56,353
  2493. I would use it to ask my mom
  2494. very hard math questions.
  2495.  
  2496. 534
  2497. 00:26:56,397 --> 00:26:58,181
  2498. Okay.
  2499. What about you, Theo?
  2500.  
  2501. 535
  2502. 00:26:58,225 --> 00:27:02,272
  2503. I would use it
  2504. for scaring people.
  2505.  
  2506. 536
  2507. 00:27:02,316 --> 00:27:04,666
  2508. All right.
  2509. So, start by smiling.
  2510.  
  2511. 537
  2512. 00:27:04,710 --> 00:27:06,625
  2513. Nice.
  2514.  
  2515. 538
  2516. 00:27:06,668 --> 00:27:09,018
  2517. Brow furrow.
  2518.  
  2519. 539
  2520. 00:27:09,062 --> 00:27:10,890
  2521. Nice one.
  2522. Eyebrow raise.
  2523.  
  2524. 540
  2525. 00:27:10,933 --> 00:27:12,587
  2526. This generation, technology
  2527.  
  2528. 541
  2529. 00:27:12,631 --> 00:27:15,068
  2530. is just surrounding them
  2531. all the time.
  2532.  
  2533. 542
  2534. 00:27:15,111 --> 00:27:17,853
  2535. It's almost like they expect
  2536. to have robots in their homes,
  2537.  
  2538. 543
  2539. 00:27:17,897 --> 00:27:22,336
  2540. and they expect these robots
  2541. to be socially intelligent.
  2542.  
  2543. 544
  2544. 00:27:22,379 --> 00:27:25,252
  2545. What makes robots smart?
  2546.  
  2547. 545
  2548. 00:27:25,295 --> 00:27:29,648
  2549. Put them in, like, a math
  2550. or biology class.
  2551.  
  2552. 546
  2553. 00:27:29,691 --> 00:27:32,259
  2554. I think you would
  2555. have to train it.
  2556.  
  2557. 547
  2558. 00:27:32,302 --> 00:27:35,218
  2559. All right.
  2560. Let's walk over here.
  2561.  
  2562. 548
  2563. 00:27:35,262 --> 00:27:37,394
  2564. So, if you smile and you
  2565. raise your eyebrows,
  2566.  
  2567. 549
  2568. 00:27:37,438 --> 00:27:39,005
  2569. it's gonna run over to you.
  2570.  
  2571. 550
  2572. 00:27:39,048 --> 00:27:40,833
  2573. Woman: It's coming over!
  2574. It's coming over! Look.
  2575.  
  2576. 551
  2577. 00:27:43,183 --> 00:27:45,272
  2578. But if you look angry,
  2579. it's gonna run away.
  2580.  
  2581. 552
  2582. 00:27:46,534 --> 00:27:48,797
  2583. -Awesome!
  2584. -Oh, that was good.
  2585.  
  2586. 553
  2587. 00:27:48,841 --> 00:27:52,366
  2588. We're training computers to read
  2589. and recognize emotions.
  2590.  
  2591. 554
  2592. 00:27:52,409 --> 00:27:53,846
  2593. Ready? Set? Go!
  2594.  
  2595. 555
  2596. 00:27:53,889 --> 00:27:57,414
  2597. And the response so far
  2598. has been really amazing.
  2599.  
  2600. 556
  2601. 00:27:57,458 --> 00:27:59,590
  2602. People are integrating this
  2603. into health apps,
  2604.  
  2605. 557
  2606. 00:27:59,634 --> 00:28:04,465
  2607. meditation apps, robots, cars.
  2608.  
  2609. 558
  2610. 00:28:04,508 --> 00:28:06,728
  2611. We're gonna see
  2612. how this unfolds.
  2613.  
  2614. 559
  2615. 00:28:06,772 --> 00:28:09,426
  2616. ♪♪
  2617.  
  2618. 560
  2619. 00:28:09,470 --> 00:28:11,602
  2620. Zilis:
  2621. Robots can contain AI,
  2622.  
  2623. 561
  2624. 00:28:11,646 --> 00:28:14,388
  2625. but the robot is just
  2626. a physical instantiation,
  2627.  
  2628. 562
  2629. 00:28:14,431 --> 00:28:16,782
  2630. and the artificial
  2631. intelligence is the brain.
  2632.  
  2633. 563
  2634. 00:28:16,825 --> 00:28:19,872
  2635. And so brains can exist purely
  2636. in software-based systems.
  2637.  
  2638. 564
  2639. 00:28:19,915 --> 00:28:22,483
  2640. They don't need to have
  2641. a physical form.
  2642.  
  2643. 565
  2644. 00:28:22,526 --> 00:28:25,094
  2645. Robots can exist without
  2646. any artificial intelligence.
  2647.  
  2648. 566
  2649. 00:28:25,138 --> 00:28:28,097
  2650. We have a lot of
  2651. dumb robots out there.
  2652.  
  2653. 567
  2654. 00:28:28,141 --> 00:28:31,753
  2655. But a dumb robot can be
  2656. a smart robot overnight,
  2657.  
  2658. 568
  2659. 00:28:31,797 --> 00:28:34,103
  2660. given the right software,
  2661. given the right sensors.
  2662.  
  2663. 569
  2664. 00:28:34,147 --> 00:28:38,629
  2665. Barrat: We can't help but impute
  2666. motive into inanimate objects.
  2667.  
  2668. 570
  2669. 00:28:38,673 --> 00:28:40,327
  2670. We do it with machines.
  2671.  
  2672. 571
  2673. 00:28:40,370 --> 00:28:41,502
  2674. We'll treat them like children.
  2675.  
  2676. 572
  2677. 00:28:41,545 --> 00:28:43,330
  2678. We'll treat them
  2679. like surrogates.
  2680.  
  2681. 573
  2682. 00:28:43,373 --> 00:28:45,027
  2683. -Goodbye!
  2684. -Goodbye!
  2685.  
  2686. 574
  2687. 00:28:45,071 --> 00:28:48,204
  2688. And we'll pay the price.
  2689.  
  2690. 575
  2691. 00:28:49,292 --> 00:28:58,998
  2692. ♪♪
  2693.  
  2694. 576
  2695. 00:28:59,041 --> 00:29:08,572
  2696. ♪♪
  2697.  
  2698. 577
  2699. 00:29:08,616 --> 00:29:10,792
  2700. Okay, welcome to ATR.
  2701.  
  2702. 578
  2703. 00:29:10,836 --> 00:29:18,060
  2704. ♪♪
  2705.  
  2706. 579
  2707. 00:29:25,067 --> 00:29:30,594
  2708. ♪♪
  2709.  
  2710. 580
  2711. 00:29:30,638 --> 00:29:36,122
  2712. ♪♪
  2713.  
  2714. 581
  2715. 00:29:47,786 --> 00:29:51,485
  2716. ♪♪
  2717.  
  2718. 582
  2719. 00:29:51,528 --> 00:29:52,791
  2720. Konnichiwa.
  2721.  
  2722. 583
  2723. 00:30:24,170 --> 00:30:29,436
  2724. ♪♪
  2725.  
  2726. 584
  2727. 00:30:53,677 --> 00:30:56,942
  2728. ♪♪
  2729.  
  2730. 585
  2731. 00:30:56,985 --> 00:30:58,682
  2732. Gourley: We build
  2733. artificial intelligence,
  2734.  
  2735. 586
  2736. 00:30:58,726 --> 00:31:02,948
  2737. and the very first thing
  2738. we want to do is replicate us.
  2739.  
  2740. 587
  2741. 00:31:02,991 --> 00:31:05,341
  2742. I think the key point will come
  2743.  
  2744. 588
  2745. 00:31:05,385 --> 00:31:09,258
  2746. when all the major senses
  2747. are replicated --
  2748.  
  2749. 589
  2750. 00:31:09,302 --> 00:31:11,130
  2751. sight...
  2752.  
  2753. 590
  2754. 00:31:11,173 --> 00:31:12,871
  2755. touch...
  2756.  
  2757. 591
  2758. 00:31:12,914 --> 00:31:14,611
  2759. smell.
  2760.  
  2761. 592
  2762. 00:31:14,655 --> 00:31:17,919
  2763. When we replicate our senses,
  2764. is that when it become alive?
  2765.  
  2766. 593
  2767. 00:31:17,963 --> 00:31:22,010
  2768. ♪♪
  2769.  
  2770. 594
  2771. 00:31:24,795 --> 00:31:27,581
  2772. ♪♪
  2773.  
  2774. 595
  2775. 00:31:27,624 --> 00:31:29,104
  2776. Nolan:
  2777. So many of our machines
  2778.  
  2779. 596
  2780. 00:31:29,148 --> 00:31:31,019
  2781. are being built
  2782. to understand us.
  2783.  
  2784. 597
  2785. 00:31:32,847 --> 00:31:34,805
  2786. But what happens when
  2787. an anthropomorphic creature
  2788.  
  2789. 598
  2790. 00:31:34,849 --> 00:31:37,417
  2791. discovers that they can
  2792. adjust their loyalty,
  2793.  
  2794. 599
  2795. 00:31:37,460 --> 00:31:40,028
  2796. adjust their courage,
  2797. adjust their avarice,
  2798.  
  2799. 600
  2800. 00:31:40,072 --> 00:31:42,291
  2801. adjust their cunning?
  2802.  
  2803. 601
  2804. 00:31:42,335 --> 00:31:44,815
  2805. ♪♪
  2806.  
  2807. 602
  2808. 00:31:44,859 --> 00:31:47,166
  2809. Musk: The average person,
  2810. they don't see killer robots
  2811.  
  2812. 603
  2813. 00:31:47,209 --> 00:31:48,645
  2814. going down the streets.
  2815.  
  2816. 604
  2817. 00:31:48,689 --> 00:31:50,996
  2818. They're like, "What are
  2819. you talking about?"
  2820.  
  2821. 605
  2822. 00:31:51,039 --> 00:31:53,955
  2823. Man, we want to make sure
  2824. that we don't have killer robots
  2825.  
  2826. 606
  2827. 00:31:53,999 --> 00:31:57,045
  2828. going down the street.
  2829.  
  2830. 607
  2831. 00:31:57,089 --> 00:31:59,439
  2832. Once they're going down
  2833. the street, it is too late.
  2834.  
  2835. 608
  2836. 00:31:59,482 --> 00:32:05,010
  2837. ♪♪
  2838.  
  2839. 609
  2840. 00:32:05,053 --> 00:32:07,099
  2841. Russell: The thing
  2842. that worries me right now,
  2843.  
  2844. 610
  2845. 00:32:07,142 --> 00:32:08,578
  2846. that keeps me awake,
  2847.  
  2848. 611
  2849. 00:32:08,622 --> 00:32:11,842
  2850. is the development
  2851. of autonomous weapons.
  2852.  
  2853. 612
  2854. 00:32:11,886 --> 00:32:19,850
  2855. ♪♪
  2856.  
  2857. 613
  2858. 00:32:19,894 --> 00:32:27,771
  2859. ♪♪
  2860.  
  2861. 614
  2862. 00:32:27,815 --> 00:32:32,733
  2863. Up to now, people have expressed
  2864. unease about drones,
  2865.  
  2866. 615
  2867. 00:32:32,776 --> 00:32:35,127
  2868. which are remotely
  2869. piloted aircraft.
  2870.  
  2871. 616
  2872. 00:32:35,170 --> 00:32:39,783
  2873. ♪♪
  2874.  
  2875. 617
  2876. 00:32:39,827 --> 00:32:43,309
  2877. If you take a drone's camera
  2878. and feed it into the AI system,
  2879.  
  2880. 618
  2881. 00:32:43,352 --> 00:32:47,443
  2882. it's a very easy step from here
  2883. to fully autonomous weapons
  2884.  
  2885. 619
  2886. 00:32:47,487 --> 00:32:50,881
  2887. that choose their own targets
  2888. and release their own missiles.
  2889.  
  2890. 620
  2891. 00:32:50,925 --> 00:32:58,150
  2892. ♪♪
  2893.  
  2894. 621
  2895. 00:32:58,193 --> 00:33:05,374
  2896. ♪♪
  2897.  
  2898. 622
  2899. 00:33:05,418 --> 00:33:12,686
  2900. ♪♪
  2901.  
  2902. 623
  2903. 00:33:12,729 --> 00:33:15,080
  2904. The expected life-span
  2905. of a human being
  2906.  
  2907. 624
  2908. 00:33:15,123 --> 00:33:16,516
  2909. in that kind of
  2910. battle environment
  2911.  
  2912. 625
  2913. 00:33:16,559 --> 00:33:20,520
  2914. would be measured in seconds.
  2915.  
  2916. 626
  2917. 00:33:20,563 --> 00:33:23,740
  2918. Singer: At one point,
  2919. drones were science fiction,
  2920.  
  2921. 627
  2922. 00:33:23,784 --> 00:33:28,832
  2923. and now they've become
  2924. the normal thing in war.
  2925.  
  2926. 628
  2927. 00:33:28,876 --> 00:33:33,402
  2928. There's over 10,000 in
  2929. U.S. military inventory alone.
  2930.  
  2931. 629
  2932. 00:33:33,446 --> 00:33:35,274
  2933. But they're not
  2934. just a U.S. phenomena.
  2935.  
  2936. 630
  2937. 00:33:35,317 --> 00:33:39,060
  2938. There's more than 80 countries
  2939. that operate them.
  2940.  
  2941. 631
  2942. 00:33:39,104 --> 00:33:41,932
  2943. Gourley: It stands to reason
  2944. that people making some
  2945.  
  2946. 632
  2947. 00:33:41,976 --> 00:33:44,587
  2948. of the most important and
  2949. difficult decisions in the world
  2950.  
  2951. 633
  2952. 00:33:44,631 --> 00:33:46,328
  2953. are gonna start to use
  2954. and implement
  2955.  
  2956. 634
  2957. 00:33:46,372 --> 00:33:48,591
  2958. artificial intelligence.
  2959.  
  2960. 635
  2961. 00:33:48,635 --> 00:33:50,724
  2962. ♪♪
  2963.  
  2964. 636
  2965. 00:33:50,767 --> 00:33:53,596
  2966. The Air Force just designed
  2967. a $400-billion jet program
  2968.  
  2969. 637
  2970. 00:33:53,640 --> 00:33:55,555
  2971. to put pilots in the sky,
  2972.  
  2973. 638
  2974. 00:33:55,598 --> 00:34:01,300
  2975. and a $500 AI, designed by
  2976. a couple of graduate students,
  2977.  
  2978. 639
  2979. 00:34:01,343 --> 00:34:03,432
  2980. is beating the best human pilots
  2981.  
  2982. 640
  2983. 00:34:03,476 --> 00:34:05,782
  2984. with a relatively
  2985. simple algorithm.
  2986.  
  2987. 641
  2988. 00:34:05,826 --> 00:34:09,395
  2989. ♪♪
  2990.  
  2991. 642
  2992. 00:34:09,438 --> 00:34:13,399
  2993. AI will have as big an impact
  2994. on the military
  2995.  
  2996. 643
  2997. 00:34:13,442 --> 00:34:17,490
  2998. as the combustion engine
  2999. had at the turn of the century.
  3000.  
  3001. 644
  3002. 00:34:17,533 --> 00:34:18,839
  3003. It will literally touch
  3004.  
  3005. 645
  3006. 00:34:18,882 --> 00:34:21,233
  3007. everything
  3008. that the military does,
  3009.  
  3010. 646
  3011. 00:34:21,276 --> 00:34:25,324
  3012. from driverless convoys
  3013. delivering logistical supplies,
  3014.  
  3015. 647
  3016. 00:34:25,367 --> 00:34:27,021
  3017. to unmanned drones
  3018.  
  3019. 648
  3020. 00:34:27,065 --> 00:34:30,764
  3021. delivering medical aid,
  3022. to computational propaganda,
  3023.  
  3024. 649
  3025. 00:34:30,807 --> 00:34:34,246
  3026. trying to win the hearts
  3027. and minds of a population.
  3028.  
  3029. 650
  3030. 00:34:34,289 --> 00:34:38,337
  3031. And so it stands to reason
  3032. that whoever has the best AI
  3033.  
  3034. 651
  3035. 00:34:38,380 --> 00:34:41,688
  3036. will probably achieve
  3037. dominance on this planet.
  3038.  
  3039. 652
  3040. 00:34:45,561 --> 00:34:47,650
  3041. At some point in
  3042. the early 21st century,
  3043.  
  3044. 653
  3045. 00:34:47,694 --> 00:34:51,219
  3046. all of mankind was
  3047. united in celebration.
  3048.  
  3049. 654
  3050. 00:34:51,263 --> 00:34:53,830
  3051. We marveled
  3052. at our own magnificence
  3053.  
  3054. 655
  3055. 00:34:53,874 --> 00:34:56,833
  3056. as we gave birth to AI.
  3057.  
  3058. 656
  3059. 00:34:56,877 --> 00:34:58,966
  3060. AI?
  3061.  
  3062. 657
  3063. 00:34:59,009 --> 00:35:00,489
  3064. You mean
  3065. artificial intelligence?
  3066.  
  3067. 658
  3068. 00:35:00,533 --> 00:35:01,751
  3069. A singular consciousness
  3070.  
  3071. 659
  3072. 00:35:01,795 --> 00:35:05,886
  3073. that spawned
  3074. an entire race of machines.
  3075.  
  3076. 660
  3077. 00:35:05,929 --> 00:35:09,716
  3078. We don't know
  3079. who struck first -- us or them,
  3080.  
  3081. 661
  3082. 00:35:09,759 --> 00:35:12,980
  3083. but we know that it was us
  3084. that scorched the sky.
  3085.  
  3086. 662
  3087. 00:35:14,677 --> 00:35:16,766
  3088. Singer: There's a long history
  3089. of science fiction,
  3090.  
  3091. 663
  3092. 00:35:16,810 --> 00:35:19,987
  3093. not just predicting the future,
  3094. but shaping the future.
  3095.  
  3096. 664
  3097. 00:35:20,030 --> 00:35:26,820
  3098. ♪♪
  3099.  
  3100. 665
  3101. 00:35:26,863 --> 00:35:30,389
  3102. Arthur Conan Doyle
  3103. writing before World War I
  3104.  
  3105. 666
  3106. 00:35:30,432 --> 00:35:34,393
  3107. on the danger of how
  3108. submarines might be used
  3109.  
  3110. 667
  3111. 00:35:34,436 --> 00:35:38,048
  3112. to carry out civilian blockades.
  3113.  
  3114. 668
  3115. 00:35:38,092 --> 00:35:40,399
  3116. At the time
  3117. he's writing this fiction,
  3118.  
  3119. 669
  3120. 00:35:40,442 --> 00:35:43,402
  3121. the Royal Navy made fun
  3122. of Arthur Conan Doyle
  3123.  
  3124. 670
  3125. 00:35:43,445 --> 00:35:45,230
  3126. for this absurd idea
  3127.  
  3128. 671
  3129. 00:35:45,273 --> 00:35:47,623
  3130. that submarines
  3131. could be useful in war.
  3132.  
  3133. 672
  3134. 00:35:47,667 --> 00:35:53,412
  3135. ♪♪
  3136.  
  3137. 673
  3138. 00:35:53,455 --> 00:35:55,370
  3139. One of the things
  3140. we've seen in history
  3141.  
  3142. 674
  3143. 00:35:55,414 --> 00:35:58,243
  3144. is that our attitude
  3145. towards technology,
  3146.  
  3147. 675
  3148. 00:35:58,286 --> 00:36:01,942
  3149. but also ethics,
  3150. are very context-dependent.
  3151.  
  3152. 676
  3153. 00:36:01,985 --> 00:36:03,726
  3154. For example, the submarine...
  3155.  
  3156. 677
  3157. 00:36:03,770 --> 00:36:06,468
  3158. nations like Great Britain
  3159. and even the United States
  3160.  
  3161. 678
  3162. 00:36:06,512 --> 00:36:09,863
  3163. found it horrifying
  3164. to use the submarine.
  3165.  
  3166. 679
  3167. 00:36:09,906 --> 00:36:13,214
  3168. In fact, the German use of the
  3169. submarine to carry out attacks
  3170.  
  3171. 680
  3172. 00:36:13,258 --> 00:36:18,480
  3173. was the reason why the United
  3174. States joined World War I.
  3175.  
  3176. 681
  3177. 00:36:18,524 --> 00:36:20,613
  3178. But move the timeline forward.
  3179.  
  3180. 682
  3181. 00:36:20,656 --> 00:36:23,529
  3182. Man: The United States
  3183. of America was suddenly
  3184.  
  3185. 683
  3186. 00:36:23,572 --> 00:36:28,403
  3187. and deliberately attacked
  3188. by the empire of Japan.
  3189.  
  3190. 684
  3191. 00:36:28,447 --> 00:36:32,190
  3192. Five hours after Pearl Harbor,
  3193. the order goes out
  3194.  
  3195. 685
  3196. 00:36:32,233 --> 00:36:36,498
  3197. to commit unrestricted
  3198. submarine warfare against Japan.
  3199.  
  3200. 686
  3201. 00:36:39,936 --> 00:36:44,289
  3202. So Arthur Conan Doyle
  3203. turned out to be right.
  3204.  
  3205. 687
  3206. 00:36:44,332 --> 00:36:46,856
  3207. Nolan: That's the great old line
  3208. about science fiction --
  3209.  
  3210. 688
  3211. 00:36:46,900 --> 00:36:48,336
  3212. It's a lie that tells the truth.
  3213.  
  3214. 689
  3215. 00:36:48,380 --> 00:36:51,470
  3216. Fellow executives,
  3217. it gives me great pleasure
  3218.  
  3219. 690
  3220. 00:36:51,513 --> 00:36:54,821
  3221. to introduce you to the future
  3222. of law enforcement...
  3223.  
  3224. 691
  3225. 00:36:54,864 --> 00:36:56,562
  3226. ED-209.
  3227.  
  3228. 692
  3229. 00:36:56,605 --> 00:37:03,612
  3230. ♪♪
  3231.  
  3232. 693
  3233. 00:37:03,656 --> 00:37:05,919
  3234. This isn't just a question
  3235. of science fiction.
  3236.  
  3237. 694
  3238. 00:37:05,962 --> 00:37:09,488
  3239. This is about what's next, about
  3240. what's happening right now.
  3241.  
  3242. 695
  3243. 00:37:09,531 --> 00:37:13,927
  3244. ♪♪
  3245.  
  3246. 696
  3247. 00:37:13,970 --> 00:37:17,496
  3248. The role of intelligent systems
  3249. is growing very rapidly
  3250.  
  3251. 697
  3252. 00:37:17,539 --> 00:37:19,324
  3253. in warfare.
  3254.  
  3255. 698
  3256. 00:37:19,367 --> 00:37:22,152
  3257. Everyone is pushing
  3258. in the unmanned realm.
  3259.  
  3260. 699
  3261. 00:37:22,196 --> 00:37:26,374
  3262. ♪♪
  3263.  
  3264. 700
  3265. 00:37:26,418 --> 00:37:28,898
  3266. Gourley: Today, the Secretary of
  3267. Defense is very, very clear --
  3268.  
  3269. 701
  3270. 00:37:28,942 --> 00:37:32,337
  3271. We will not create fully
  3272. autonomous attacking vehicles.
  3273.  
  3274. 702
  3275. 00:37:32,380 --> 00:37:34,643
  3276. Not everyone
  3277. is gonna hold themselves
  3278.  
  3279. 703
  3280. 00:37:34,687 --> 00:37:36,515
  3281. to that same set of values.
  3282.  
  3283. 704
  3284. 00:37:36,558 --> 00:37:40,693
  3285. And when China and Russia start
  3286. deploying autonomous vehicles
  3287.  
  3288. 705
  3289. 00:37:40,736 --> 00:37:45,611
  3290. that can attack and kill, what's
  3291. the move that we're gonna make?
  3292.  
  3293. 706
  3294. 00:37:45,654 --> 00:37:49,963
  3295. ♪♪
  3296.  
  3297. 707
  3298. 00:37:50,006 --> 00:37:51,617
  3299. Russell: You can't say,
  3300. "Well, we're gonna use
  3301.  
  3302. 708
  3303. 00:37:51,660 --> 00:37:53,967
  3304. autonomous weapons
  3305. for our military dominance,
  3306.  
  3307. 709
  3308. 00:37:54,010 --> 00:37:56,796
  3309. but no one else
  3310. is gonna use them."
  3311.  
  3312. 710
  3313. 00:37:56,839 --> 00:38:00,495
  3314. If you make these weapons,
  3315. they're gonna be used to attack
  3316.  
  3317. 711
  3318. 00:38:00,539 --> 00:38:03,324
  3319. human populations
  3320. in large numbers.
  3321.  
  3322. 712
  3323. 00:38:03,368 --> 00:38:12,507
  3324. ♪♪
  3325.  
  3326. 713
  3327. 00:38:12,551 --> 00:38:14,596
  3328. Autonomous weapons are,
  3329. by their nature,
  3330.  
  3331. 714
  3332. 00:38:14,640 --> 00:38:16,468
  3333. weapons of mass destruction,
  3334.  
  3335. 715
  3336. 00:38:16,511 --> 00:38:19,862
  3337. because it doesn't need a human
  3338. being to guide it or carry it.
  3339.  
  3340. 716
  3341. 00:38:19,906 --> 00:38:22,517
  3342. You only need one person,
  3343. to, you know,
  3344.  
  3345. 717
  3346. 00:38:22,561 --> 00:38:25,781
  3347. write a little program.
  3348.  
  3349. 718
  3350. 00:38:25,825 --> 00:38:30,220
  3351. It just captures
  3352. the complexity of this field.
  3353.  
  3354. 719
  3355. 00:38:30,264 --> 00:38:32,571
  3356. It is cool.
  3357. It is important.
  3358.  
  3359. 720
  3360. 00:38:32,614 --> 00:38:34,573
  3361. It is amazing.
  3362.  
  3363. 721
  3364. 00:38:34,616 --> 00:38:37,053
  3365. It is also frightening.
  3366.  
  3367. 722
  3368. 00:38:37,097 --> 00:38:38,968
  3369. And it's all about trust.
  3370.  
  3371. 723
  3372. 00:38:42,102 --> 00:38:44,583
  3373. It's an open letter about
  3374. artificial intelligence,
  3375.  
  3376. 724
  3377. 00:38:44,626 --> 00:38:47,063
  3378. signed by some of
  3379. the biggest names in science.
  3380.  
  3381. 725
  3382. 00:38:47,107 --> 00:38:48,413
  3383. What do they want?
  3384.  
  3385. 726
  3386. 00:38:48,456 --> 00:38:50,763
  3387. Ban the use of
  3388. autonomous weapons.
  3389.  
  3390. 727
  3391. 00:38:50,806 --> 00:38:52,373
  3392. Woman: The author stated,
  3393.  
  3394. 728
  3395. 00:38:52,417 --> 00:38:54,375
  3396. "Autonomous weapons
  3397. have been described
  3398.  
  3399. 729
  3400. 00:38:54,419 --> 00:38:56,595
  3401. as the third revolution
  3402. in warfare."
  3403.  
  3404. 730
  3405. 00:38:56,638 --> 00:38:58,553
  3406. Woman #2: ...thousand
  3407. artificial-intelligence
  3408. specialists
  3409.  
  3410. 731
  3411. 00:38:58,597 --> 00:39:01,817
  3412. calling for a global ban
  3413. on killer robots.
  3414.  
  3415. 732
  3416. 00:39:01,861 --> 00:39:04,342
  3417. Tegmark:
  3418. This open letter basically says
  3419.  
  3420. 733
  3421. 00:39:04,385 --> 00:39:06,344
  3422. that we should redefine the goal
  3423.  
  3424. 734
  3425. 00:39:06,387 --> 00:39:07,954
  3426. of the field of
  3427. artificial intelligence
  3428.  
  3429. 735
  3430. 00:39:07,997 --> 00:39:11,610
  3431. away from just creating pure,
  3432. undirected intelligence,
  3433.  
  3434. 736
  3435. 00:39:11,653 --> 00:39:13,655
  3436. towards creating
  3437. beneficial intelligence.
  3438.  
  3439. 737
  3440. 00:39:13,699 --> 00:39:16,092
  3441. The development of AI
  3442. is not going to stop.
  3443.  
  3444. 738
  3445. 00:39:16,136 --> 00:39:18,094
  3446. It is going to continue
  3447. and get better.
  3448.  
  3449. 739
  3450. 00:39:18,138 --> 00:39:19,835
  3451. If the international community
  3452.  
  3453. 740
  3454. 00:39:19,879 --> 00:39:21,968
  3455. isn't putting
  3456. certain controls on this,
  3457.  
  3458. 741
  3459. 00:39:22,011 --> 00:39:24,666
  3460. people will develop things
  3461. that can do anything.
  3462.  
  3463. 742
  3464. 00:39:24,710 --> 00:39:27,365
  3465. Woman: The letter says
  3466. that we are years, not decades,
  3467.  
  3468. 743
  3469. 00:39:27,408 --> 00:39:28,714
  3470. away from these weapons
  3471. being deployed.
  3472.  
  3473. 744
  3474. 00:39:28,757 --> 00:39:30,106
  3475. So first of all...
  3476.  
  3477. 745
  3478. 00:39:30,150 --> 00:39:32,413
  3479. We had 6,000 signatories
  3480. of that letter,
  3481.  
  3482. 746
  3483. 00:39:32,457 --> 00:39:35,155
  3484. including many of
  3485. the major figures in the field.
  3486.  
  3487. 747
  3488. 00:39:37,026 --> 00:39:39,942
  3489. I'm getting a lot of visits
  3490. from high-ranking officials
  3491.  
  3492. 748
  3493. 00:39:39,986 --> 00:39:42,989
  3494. who wish to emphasize that
  3495. American military dominance
  3496.  
  3497. 749
  3498. 00:39:43,032 --> 00:39:45,731
  3499. is very important,
  3500. and autonomous weapons
  3501.  
  3502. 750
  3503. 00:39:45,774 --> 00:39:50,083
  3504. may be part of
  3505. the Defense Department's plan.
  3506.  
  3507. 751
  3508. 00:39:50,126 --> 00:39:52,433
  3509. That's very, very scary,
  3510. because a value system
  3511.  
  3512. 752
  3513. 00:39:52,477 --> 00:39:54,479
  3514. of military developers
  3515. of technology
  3516.  
  3517. 753
  3518. 00:39:54,522 --> 00:39:57,307
  3519. is not the same as a value
  3520. system of the human race.
  3521.  
  3522. 754
  3523. 00:39:57,351 --> 00:40:00,746
  3524. ♪♪
  3525.  
  3526. 755
  3527. 00:40:00,789 --> 00:40:02,922
  3528. Markoff: Out of the concerns
  3529. about the possibility
  3530.  
  3531. 756
  3532. 00:40:02,965 --> 00:40:06,665
  3533. that this technology might be
  3534. a threat to human existence,
  3535.  
  3536. 757
  3537. 00:40:06,708 --> 00:40:08,144
  3538. a number of the technologists
  3539.  
  3540. 758
  3541. 00:40:08,188 --> 00:40:09,972
  3542. have funded
  3543. the Future of Life Institute
  3544.  
  3545. 759
  3546. 00:40:10,016 --> 00:40:12,192
  3547. to try to grapple
  3548. with these problems.
  3549.  
  3550. 760
  3551. 00:40:13,193 --> 00:40:14,847
  3552. All of these guys are secretive,
  3553.  
  3554. 761
  3555. 00:40:14,890 --> 00:40:16,805
  3556. and so it's interesting
  3557. to me to see them,
  3558.  
  3559. 762
  3560. 00:40:16,849 --> 00:40:20,635
  3561. you know, all together.
  3562.  
  3563. 763
  3564. 00:40:20,679 --> 00:40:24,030
  3565. Everything we have is a result
  3566. of our intelligence.
  3567.  
  3568. 764
  3569. 00:40:24,073 --> 00:40:26,641
  3570. It's not the result
  3571. of our big, scary teeth
  3572.  
  3573. 765
  3574. 00:40:26,685 --> 00:40:29,470
  3575. or our large claws
  3576. or our enormous muscles.
  3577.  
  3578. 766
  3579. 00:40:29,514 --> 00:40:32,473
  3580. It's because we're actually
  3581. relatively intelligent.
  3582.  
  3583. 767
  3584. 00:40:32,517 --> 00:40:35,520
  3585. And among my generation,
  3586. we're all having
  3587.  
  3588. 768
  3589. 00:40:35,563 --> 00:40:37,086
  3590. what we call "holy cow,"
  3591.  
  3592. 769
  3593. 00:40:37,130 --> 00:40:39,045
  3594. or "holy something else"
  3595. moments,
  3596.  
  3597. 770
  3598. 00:40:39,088 --> 00:40:41,003
  3599. because we see
  3600. that the technology
  3601.  
  3602. 771
  3603. 00:40:41,047 --> 00:40:44,180
  3604. is accelerating faster
  3605. than we expected.
  3606.  
  3607. 772
  3608. 00:40:44,224 --> 00:40:46,705
  3609. I remember sitting
  3610. around the table there
  3611.  
  3612. 773
  3613. 00:40:46,748 --> 00:40:50,099
  3614. with some of the best and
  3615. the smartest minds in the world,
  3616.  
  3617. 774
  3618. 00:40:50,143 --> 00:40:52,058
  3619. and what really
  3620. struck me was,
  3621.  
  3622. 775
  3623. 00:40:52,101 --> 00:40:56,149
  3624. maybe the human brain
  3625. is not able to fully grasp
  3626.  
  3627. 776
  3628. 00:40:56,192 --> 00:40:58,673
  3629. the complexity of the world
  3630. that we're confronted with.
  3631.  
  3632. 777
  3633. 00:40:58,717 --> 00:41:01,415
  3634. Russell:
  3635. As it's currently constructed,
  3636.  
  3637. 778
  3638. 00:41:01,459 --> 00:41:04,766
  3639. the road that AI is following
  3640. heads off a cliff,
  3641.  
  3642. 779
  3643. 00:41:04,810 --> 00:41:07,595
  3644. and we need to change
  3645. the direction that we're going
  3646.  
  3647. 780
  3648. 00:41:07,639 --> 00:41:10,729
  3649. so that we don't take
  3650. the human race off the cliff.
  3651.  
  3652. 781
  3653. 00:41:13,558 --> 00:41:17,126
  3654. Musk: Google acquired DeepMind
  3655. several years ago.
  3656.  
  3657. 782
  3658. 00:41:17,170 --> 00:41:18,737
  3659. DeepMind operates
  3660.  
  3661. 783
  3662. 00:41:18,780 --> 00:41:22,088
  3663. as a semi-independent
  3664. subsidiary of Google.
  3665.  
  3666. 784
  3667. 00:41:22,131 --> 00:41:24,960
  3668. The thing that makes
  3669. DeepMind unique
  3670.  
  3671. 785
  3672. 00:41:25,004 --> 00:41:26,919
  3673. is that DeepMind
  3674. is absolutely focused
  3675.  
  3676. 786
  3677. 00:41:26,962 --> 00:41:30,313
  3678. on creating digital
  3679. superintelligence --
  3680.  
  3681. 787
  3682. 00:41:30,357 --> 00:41:34,056
  3683. an AI that is vastly smarter
  3684. than any human on Earth
  3685.  
  3686. 788
  3687. 00:41:34,100 --> 00:41:36,624
  3688. and ultimately smarter than
  3689. all humans on Earth combined.
  3690.  
  3691. 789
  3692. 00:41:36,668 --> 00:41:40,715
  3693. This is from the DeepMind
  3694. reinforcement learning system.
  3695.  
  3696. 790
  3697. 00:41:40,759 --> 00:41:43,544
  3698. Basically wakes up
  3699. like a newborn baby
  3700.  
  3701. 791
  3702. 00:41:43,588 --> 00:41:46,852
  3703. and is shown the screen
  3704. of an Atari video game
  3705.  
  3706. 792
  3707. 00:41:46,895 --> 00:41:50,508
  3708. and then has to learn
  3709. to play the video game.
  3710.  
  3711. 793
  3712. 00:41:50,551 --> 00:41:55,600
  3713. It knows nothing about objects,
  3714. about motion, about time.
  3715.  
  3716. 794
  3717. 00:41:57,602 --> 00:41:59,604
  3718. It only knows that there's
  3719. an image on the screen
  3720.  
  3721. 795
  3722. 00:41:59,647 --> 00:42:02,563
  3723. and there's a score.
  3724.  
  3725. 796
  3726. 00:42:02,607 --> 00:42:06,436
  3727. So, if your baby woke up
  3728. the day it was born
  3729.  
  3730. 797
  3731. 00:42:06,480 --> 00:42:08,090
  3732. and, by late afternoon,
  3733.  
  3734. 798
  3735. 00:42:08,134 --> 00:42:11,093
  3736. was playing
  3737. 40 different Atari video games
  3738.  
  3739. 799
  3740. 00:42:11,137 --> 00:42:15,315
  3741. at a superhuman level,
  3742. you would be terrified.
  3743.  
  3744. 800
  3745. 00:42:15,358 --> 00:42:19,101
  3746. You would say, "My baby
  3747. is possessed. Send it back."
  3748.  
  3749. 801
  3750. 00:42:19,145 --> 00:42:23,584
  3751. Musk: The DeepMind system
  3752. can win at any game.
  3753.  
  3754. 802
  3755. 00:42:23,628 --> 00:42:27,588
  3756. It can already beat all
  3757. the original Atari games.
  3758.  
  3759. 803
  3760. 00:42:27,632 --> 00:42:29,155
  3761. It is superhuman.
  3762.  
  3763. 804
  3764. 00:42:29,198 --> 00:42:31,636
  3765. It plays the games at superspeed
  3766. in less than a minute.
  3767.  
  3768. 805
  3769. 00:42:35,640 --> 00:42:37,032
  3770. ♪♪
  3771.  
  3772. 806
  3773. 00:42:37,076 --> 00:42:38,643
  3774. DeepMind turned
  3775. to another challenge,
  3776.  
  3777. 807
  3778. 00:42:38,686 --> 00:42:40,558
  3779. and the challenge
  3780. was the game of Go,
  3781.  
  3782. 808
  3783. 00:42:40,601 --> 00:42:42,603
  3784. which people
  3785. have generally argued
  3786.  
  3787. 809
  3788. 00:42:42,647 --> 00:42:45,084
  3789. has been beyond
  3790. the power of computers
  3791.  
  3792. 810
  3793. 00:42:45,127 --> 00:42:48,304
  3794. to play with
  3795. the best human Go players.
  3796.  
  3797. 811
  3798. 00:42:48,348 --> 00:42:51,264
  3799. First, they challenged
  3800. a European Go champion.
  3801.  
  3802. 812
  3803. 00:42:53,222 --> 00:42:55,834
  3804. Then they challenged
  3805. a Korean Go champion.
  3806.  
  3807. 813
  3808. 00:42:55,877 --> 00:42:57,836
  3809. Man:
  3810. Please start the game.
  3811.  
  3812. 814
  3813. 00:42:57,879 --> 00:42:59,838
  3814. And they were able
  3815. to win both times
  3816.  
  3817. 815
  3818. 00:42:59,881 --> 00:43:02,797
  3819. in kind of striking fashion.
  3820.  
  3821. 816
  3822. 00:43:02,841 --> 00:43:05,017
  3823. Nolan: You were reading articles
  3824. in New York Timesyears ago
  3825.  
  3826. 817
  3827. 00:43:05,060 --> 00:43:09,761
  3828. talking about how Go would take
  3829. 100 years for us to solve.
  3830.  
  3831. 818
  3832. 00:43:09,804 --> 00:43:11,110
  3833. Urban:
  3834. People said, "Well, you know,
  3835.  
  3836. 819
  3837. 00:43:11,153 --> 00:43:13,460
  3838. but that's still just a board.
  3839.  
  3840. 820
  3841. 00:43:13,503 --> 00:43:15,027
  3842. Poker is an art.
  3843.  
  3844. 821
  3845. 00:43:15,070 --> 00:43:16,419
  3846. Poker involves reading people.
  3847.  
  3848. 822
  3849. 00:43:16,463 --> 00:43:18,073
  3850. Poker involves lying
  3851. and bluffing.
  3852.  
  3853. 823
  3854. 00:43:18,117 --> 00:43:19,553
  3855. It's not an exact thing.
  3856.  
  3857. 824
  3858. 00:43:19,597 --> 00:43:21,381
  3859. That will never be,
  3860. you know, a computer.
  3861.  
  3862. 825
  3863. 00:43:21,424 --> 00:43:22,861
  3864. You can't do that."
  3865.  
  3866. 826
  3867. 00:43:22,904 --> 00:43:24,732
  3868. They took the best
  3869. poker players in the world,
  3870.  
  3871. 827
  3872. 00:43:24,776 --> 00:43:27,387
  3873. and it took seven days
  3874. for the computer
  3875.  
  3876. 828
  3877. 00:43:27,430 --> 00:43:30,520
  3878. to start demolishing the humans.
  3879.  
  3880. 829
  3881. 00:43:30,564 --> 00:43:32,261
  3882. So it's the best poker player
  3883. in the world,
  3884.  
  3885. 830
  3886. 00:43:32,305 --> 00:43:34,655
  3887. it's the best Go player in the
  3888. world, and the pattern here
  3889.  
  3890. 831
  3891. 00:43:34,699 --> 00:43:37,440
  3892. is that AI might take
  3893. a little while
  3894.  
  3895. 832
  3896. 00:43:37,484 --> 00:43:40,443
  3897. to wrap its tentacles
  3898. around a new skill,
  3899.  
  3900. 833
  3901. 00:43:40,487 --> 00:43:44,883
  3902. but when it does, when it
  3903. gets it, it is unstoppable.
  3904.  
  3905. 834
  3906. 00:43:44,926 --> 00:43:51,977
  3907. ♪♪
  3908.  
  3909. 835
  3910. 00:43:52,020 --> 00:43:55,110
  3911. DeepMind's AI has
  3912. administrator-level access
  3913.  
  3914. 836
  3915. 00:43:55,154 --> 00:43:57,156
  3916. to Google's servers
  3917.  
  3918. 837
  3919. 00:43:57,199 --> 00:44:00,768
  3920. to optimize energy usage
  3921. at the data centers.
  3922.  
  3923. 838
  3924. 00:44:00,812 --> 00:44:04,816
  3925. However, this could be
  3926. an unintentional Trojan horse.
  3927.  
  3928. 839
  3929. 00:44:04,859 --> 00:44:07,253
  3930. DeepMind has to have complete
  3931. control of the data centers,
  3932.  
  3933. 840
  3934. 00:44:07,296 --> 00:44:08,950
  3935. so with a little
  3936. software update,
  3937.  
  3938. 841
  3939. 00:44:08,994 --> 00:44:10,691
  3940. that AI could take
  3941. complete control
  3942.  
  3943. 842
  3944. 00:44:10,735 --> 00:44:12,214
  3945. of the whole Google system,
  3946.  
  3947. 843
  3948. 00:44:12,258 --> 00:44:13,607
  3949. which means
  3950. they can do anything.
  3951.  
  3952. 844
  3953. 00:44:13,651 --> 00:44:14,913
  3954. They could look
  3955. at all your data.
  3956.  
  3957. 845
  3958. 00:44:14,956 --> 00:44:16,131
  3959. They could do anything.
  3960.  
  3961. 846
  3962. 00:44:16,175 --> 00:44:18,917
  3963. ♪♪
  3964.  
  3965. 847
  3966. 00:44:20,135 --> 00:44:23,051
  3967. We're rapidly heading towards
  3968. digital superintelligence
  3969.  
  3970. 848
  3971. 00:44:23,095 --> 00:44:24,313
  3972. that far exceeds any human.
  3973.  
  3974. 849
  3975. 00:44:24,357 --> 00:44:26,402
  3976. I think it's very obvious.
  3977.  
  3978. 850
  3979. 00:44:26,446 --> 00:44:27,708
  3980. Barrat:
  3981. The problem is, we're not gonna
  3982.  
  3983. 851
  3984. 00:44:27,752 --> 00:44:29,710
  3985. suddenly hit
  3986. human-level intelligence
  3987.  
  3988. 852
  3989. 00:44:29,754 --> 00:44:33,105
  3990. and say,
  3991. "Okay, let's stop research."
  3992.  
  3993. 853
  3994. 00:44:33,148 --> 00:44:34,715
  3995. It's gonna go beyond
  3996. human-level intelligence
  3997.  
  3998. 854
  3999. 00:44:34,759 --> 00:44:36,195
  4000. into what's called
  4001. "superintelligence,"
  4002.  
  4003. 855
  4004. 00:44:36,238 --> 00:44:39,459
  4005. and that's anything
  4006. smarter than us.
  4007.  
  4008. 856
  4009. 00:44:39,502 --> 00:44:41,287
  4010. Tegmark:
  4011. AI at the superhuman level,
  4012.  
  4013. 857
  4014. 00:44:41,330 --> 00:44:42,810
  4015. if we succeed with that,
  4016. will be
  4017.  
  4018. 858
  4019. 00:44:42,854 --> 00:44:46,553
  4020. by far the most powerful
  4021. invention we've ever made
  4022.  
  4023. 859
  4024. 00:44:46,596 --> 00:44:50,296
  4025. and the last invention
  4026. we ever have to make.
  4027.  
  4028. 860
  4029. 00:44:50,339 --> 00:44:53,168
  4030. And if we create AI
  4031. that's smarter than us,
  4032.  
  4033. 861
  4034. 00:44:53,212 --> 00:44:54,735
  4035. we have to be open
  4036. to the possibility
  4037.  
  4038. 862
  4039. 00:44:54,779 --> 00:44:57,520
  4040. that we might actually
  4041. lose control to them.
  4042.  
  4043. 863
  4044. 00:44:57,564 --> 00:45:00,741
  4045. ♪♪
  4046.  
  4047. 864
  4048. 00:45:00,785 --> 00:45:02,612
  4049. Russell: Let's say
  4050. you give it some objective,
  4051.  
  4052. 865
  4053. 00:45:02,656 --> 00:45:04,745
  4054. like curing cancer,
  4055. and then you discover
  4056.  
  4057. 866
  4058. 00:45:04,789 --> 00:45:06,965
  4059. that the way
  4060. it chooses to go about that
  4061.  
  4062. 867
  4063. 00:45:07,008 --> 00:45:08,444
  4064. is actually in conflict
  4065.  
  4066. 868
  4067. 00:45:08,488 --> 00:45:12,405
  4068. with a lot of other things
  4069. you care about.
  4070.  
  4071. 869
  4072. 00:45:12,448 --> 00:45:16,496
  4073. Musk: AI doesn't have to be evil
  4074. to destroy humanity.
  4075.  
  4076. 870
  4077. 00:45:16,539 --> 00:45:20,674
  4078. If AI has a goal, and humanity
  4079. just happens to be in the way,
  4080.  
  4081. 871
  4082. 00:45:20,718 --> 00:45:22,894
  4083. it will destroy humanity
  4084. as a matter of course,
  4085.  
  4086. 872
  4087. 00:45:22,937 --> 00:45:25,113
  4088. without even thinking about it.
  4089. No hard feelings.
  4090.  
  4091. 873
  4092. 00:45:25,157 --> 00:45:27,072
  4093. It's just like
  4094. if we're building a road
  4095.  
  4096. 874
  4097. 00:45:27,115 --> 00:45:29,770
  4098. and an anthill happens
  4099. to be in the way...
  4100.  
  4101. 875
  4102. 00:45:29,814 --> 00:45:31,467
  4103. We don't hate ants.
  4104.  
  4105. 876
  4106. 00:45:31,511 --> 00:45:33,165
  4107. We're just building a road.
  4108.  
  4109. 877
  4110. 00:45:33,208 --> 00:45:34,557
  4111. And so goodbye, anthill.
  4112.  
  4113. 878
  4114. 00:45:34,601 --> 00:45:37,952
  4115. ♪♪
  4116.  
  4117. 879
  4118. 00:45:37,996 --> 00:45:40,172
  4119. It's tempting
  4120. to dismiss these concerns,
  4121.  
  4122. 880
  4123. 00:45:40,215 --> 00:45:42,783
  4124. 'cause it's, like,
  4125. something that might happen
  4126.  
  4127. 881
  4128. 00:45:42,827 --> 00:45:47,396
  4129. in a few decades or 100 years,
  4130. so why worry?
  4131.  
  4132. 882
  4133. 00:45:47,440 --> 00:45:50,704
  4134. Russell: But if you go back
  4135. to September 11, 1933,
  4136.  
  4137. 883
  4138. 00:45:50,748 --> 00:45:52,401
  4139. Ernest Rutherford,
  4140.  
  4141. 884
  4142. 00:45:52,445 --> 00:45:54,795
  4143. who was the most well-known
  4144. nuclear physicist of his time,
  4145.  
  4146. 885
  4147. 00:45:54,839 --> 00:45:56,318
  4148. said that the possibility
  4149.  
  4150. 886
  4151. 00:45:56,362 --> 00:45:58,668
  4152. of ever extracting
  4153. useful amounts of energy
  4154.  
  4155. 887
  4156. 00:45:58,712 --> 00:46:00,801
  4157. from the transmutation
  4158. of atoms, as he called it,
  4159.  
  4160. 888
  4161. 00:46:00,845 --> 00:46:03,151
  4162. was moonshine.
  4163.  
  4164. 889
  4165. 00:46:03,195 --> 00:46:04,849
  4166. The next morning, Leo Szilard,
  4167.  
  4168. 890
  4169. 00:46:04,892 --> 00:46:06,502
  4170. who was a much
  4171. younger physicist,
  4172.  
  4173. 891
  4174. 00:46:06,546 --> 00:46:09,984
  4175. read this and got really annoyed
  4176. and figured out
  4177.  
  4178. 892
  4179. 00:46:10,028 --> 00:46:11,943
  4180. how to make
  4181. a nuclear chain reaction
  4182.  
  4183. 893
  4184. 00:46:11,986 --> 00:46:13,379
  4185. just a few months later.
  4186.  
  4187. 894
  4188. 00:46:13,422 --> 00:46:20,560
  4189. ♪♪
  4190.  
  4191. 895
  4192. 00:46:20,603 --> 00:46:23,693
  4193. We have spent more
  4194. than $2 billion
  4195.  
  4196. 896
  4197. 00:46:23,737 --> 00:46:27,523
  4198. on the greatest
  4199. scientific gamble in history.
  4200.  
  4201. 897
  4202. 00:46:27,567 --> 00:46:30,222
  4203. Russell: So when people say
  4204. that, "Oh, this is so far off
  4205.  
  4206. 898
  4207. 00:46:30,265 --> 00:46:32,528
  4208. in the future, we don't have
  4209. to worry about it,"
  4210.  
  4211. 899
  4212. 00:46:32,572 --> 00:46:36,271
  4213. it might only be three, four
  4214. breakthroughs of that magnitude
  4215.  
  4216. 900
  4217. 00:46:36,315 --> 00:46:40,275
  4218. that will get us from here
  4219. to superintelligent machines.
  4220.  
  4221. 901
  4222. 00:46:40,319 --> 00:46:42,974
  4223. Tegmark: If it's gonna take
  4224. 20 years to figure out
  4225.  
  4226. 902
  4227. 00:46:43,017 --> 00:46:45,237
  4228. how to keep AI beneficial,
  4229.  
  4230. 903
  4231. 00:46:45,280 --> 00:46:48,849
  4232. then we should start today,
  4233. not at the last second
  4234.  
  4235. 904
  4236. 00:46:48,893 --> 00:46:51,460
  4237. when some dudes
  4238. drinking Red Bull
  4239.  
  4240. 905
  4241. 00:46:51,504 --> 00:46:53,332
  4242. decide to flip the switch
  4243. and test the thing.
  4244.  
  4245. 906
  4246. 00:46:53,375 --> 00:46:56,770
  4247. ♪♪
  4248.  
  4249. 907
  4250. 00:46:56,814 --> 00:46:58,859
  4251. Musk:
  4252. We have five years.
  4253.  
  4254. 908
  4255. 00:46:58,903 --> 00:47:00,600
  4256. I think
  4257. digital superintelligence
  4258.  
  4259. 909
  4260. 00:47:00,643 --> 00:47:03,864
  4261. will happen in my lifetime.
  4262.  
  4263. 910
  4264. 00:47:03,908 --> 00:47:05,735
  4265. 100%.
  4266.  
  4267. 911
  4268. 00:47:05,779 --> 00:47:07,215
  4269. Barrat: When this happens,
  4270.  
  4271. 912
  4272. 00:47:07,259 --> 00:47:09,696
  4273. it will be surrounded
  4274. by a bunch of people
  4275.  
  4276. 913
  4277. 00:47:09,739 --> 00:47:13,091
  4278. who are really just excited
  4279. about the technology.
  4280.  
  4281. 914
  4282. 00:47:13,134 --> 00:47:15,571
  4283. They want to see it succeed,
  4284. but they're not anticipating
  4285.  
  4286. 915
  4287. 00:47:15,615 --> 00:47:16,964
  4288. that it can get out of control.
  4289.  
  4290. 916
  4291. 00:47:17,008 --> 00:47:24,450
  4292. ♪♪
  4293.  
  4294. 917
  4295. 00:47:25,494 --> 00:47:28,584
  4296. Oh, my God, I trust
  4297. my computer so much.
  4298.  
  4299. 918
  4300. 00:47:28,628 --> 00:47:30,195
  4301. That's an amazing question.
  4302.  
  4303. 919
  4304. 00:47:30,238 --> 00:47:31,457
  4305. I don't trust
  4306. my computer.
  4307.  
  4308. 920
  4309. 00:47:31,500 --> 00:47:32,937
  4310. If it's on,
  4311. I take it off.
  4312.  
  4313. 921
  4314. 00:47:32,980 --> 00:47:34,242
  4315. Like, even when it's off,
  4316.  
  4317. 922
  4318. 00:47:34,286 --> 00:47:35,896
  4319. I still think it's on.
  4320. Like, you know?
  4321.  
  4322. 923
  4323. 00:47:35,940 --> 00:47:37,637
  4324. Like, you really cannot tru--
  4325. Like, the webcams,
  4326.  
  4327. 924
  4328. 00:47:37,680 --> 00:47:39,595
  4329. you don't know if, like,
  4330. someone might turn it...
  4331.  
  4332. 925
  4333. 00:47:39,639 --> 00:47:41,249
  4334. You don't know, like.
  4335.  
  4336. 926
  4337. 00:47:41,293 --> 00:47:42,903
  4338. I don't trust my computer.
  4339.  
  4340. 927
  4341. 00:47:42,947 --> 00:47:46,907
  4342. Like, in my phone,
  4343. every time they ask me
  4344.  
  4345. 928
  4346. 00:47:46,951 --> 00:47:49,475
  4347. "Can we send your
  4348. information to Apple?"
  4349.  
  4350. 929
  4351. 00:47:49,518 --> 00:47:50,998
  4352. every time, I...
  4353.  
  4354. 930
  4355. 00:47:51,042 --> 00:47:53,087
  4356. So, I don't trust my phone.
  4357.  
  4358. 931
  4359. 00:47:53,131 --> 00:47:56,743
  4360. Okay. So, part of it is,
  4361. yes, I do trust it,
  4362.  
  4363. 932
  4364. 00:47:56,786 --> 00:48:00,660
  4365. because it would be really
  4366. hard to get through the day
  4367.  
  4368. 933
  4369. 00:48:00,703 --> 00:48:04,011
  4370. in the way our world is
  4371. set up without computers.
  4372.  
  4373. 934
  4374. 00:48:04,055 --> 00:48:05,360
  4375. ♪♪
  4376.  
  4377. 935
  4378. 00:48:10,975 --> 00:48:13,368
  4379. Dr. Herman: Trust is
  4380. such a human experience.
  4381.  
  4382. 936
  4383. 00:48:13,412 --> 00:48:21,246
  4384. ♪♪
  4385.  
  4386. 937
  4387. 00:48:21,289 --> 00:48:25,119
  4388. I have a patient coming in
  4389. with an intracranial aneurysm.
  4390.  
  4391. 938
  4392. 00:48:25,163 --> 00:48:29,994
  4393. ♪♪
  4394.  
  4395. 939
  4396. 00:48:30,037 --> 00:48:31,691
  4397. They want to look
  4398. in my eyes and know
  4399.  
  4400. 940
  4401. 00:48:31,734 --> 00:48:34,955
  4402. that they can trust
  4403. this person with their life.
  4404.  
  4405. 941
  4406. 00:48:34,999 --> 00:48:39,394
  4407. I'm not horribly concerned
  4408. about anything.
  4409.  
  4410. 942
  4411. 00:48:39,438 --> 00:48:40,830
  4412. Good.
  4413. Part of that
  4414.  
  4415. 943
  4416. 00:48:40,874 --> 00:48:42,920
  4417. is because
  4418. I have confidence in you.
  4419.  
  4420. 944
  4421. 00:48:42,963 --> 00:48:50,710
  4422. ♪♪
  4423.  
  4424. 945
  4425. 00:48:50,753 --> 00:48:52,233
  4426. This procedure
  4427. we're doing today
  4428.  
  4429. 946
  4430. 00:48:52,277 --> 00:48:57,151
  4431. 20 years ago
  4432. was essentially impossible.
  4433.  
  4434. 947
  4435. 00:48:57,195 --> 00:49:00,328
  4436. We just didn't have the
  4437. materials and the technologies.
  4438.  
  4439. 948
  4440. 00:49:04,202 --> 00:49:13,385
  4441. ♪♪
  4442.  
  4443. 949
  4444. 00:49:13,428 --> 00:49:22,655
  4445. ♪♪
  4446.  
  4447. 950
  4448. 00:49:22,698 --> 00:49:26,485
  4449. So, the coil is barely
  4450. in there right now.
  4451.  
  4452. 951
  4453. 00:49:26,528 --> 00:49:29,923
  4454. It's just a feather
  4455. holding it in.
  4456.  
  4457. 952
  4458. 00:49:29,967 --> 00:49:32,012
  4459. It's nervous time.
  4460.  
  4461. 953
  4462. 00:49:32,056 --> 00:49:36,147
  4463. ♪♪
  4464.  
  4465. 954
  4466. 00:49:36,190 --> 00:49:37,626
  4467. We're just in purgatory,
  4468.  
  4469. 955
  4470. 00:49:37,670 --> 00:49:40,673
  4471. intellectual,
  4472. humanistic purgatory,
  4473.  
  4474. 956
  4475. 00:49:40,716 --> 00:49:43,632
  4476. and AI might know
  4477. exactly what to do here.
  4478.  
  4479. 957
  4480. 00:49:43,676 --> 00:49:50,596
  4481. ♪♪
  4482.  
  4483. 958
  4484. 00:49:50,639 --> 00:49:52,554
  4485. We've got the coil
  4486. into the aneurysm.
  4487.  
  4488. 959
  4489. 00:49:52,598 --> 00:49:54,556
  4490. But it wasn't in
  4491. tremendously well
  4492.  
  4493. 960
  4494. 00:49:54,600 --> 00:49:56,428
  4495. that I knew that it would stay,
  4496.  
  4497. 961
  4498. 00:49:56,471 --> 00:50:01,041
  4499. so with a maybe 20% risk
  4500. of a very bad situation,
  4501.  
  4502. 962
  4503. 00:50:01,085 --> 00:50:04,436
  4504. I elected
  4505. to just bring her back.
  4506.  
  4507. 963
  4508. 00:50:04,479 --> 00:50:05,959
  4509. Because of my relationship
  4510. with her
  4511.  
  4512. 964
  4513. 00:50:06,003 --> 00:50:08,222
  4514. and knowing the difficulties
  4515. of coming in
  4516.  
  4517. 965
  4518. 00:50:08,266 --> 00:50:11,051
  4519. and having the procedure,
  4520. I consider things,
  4521.  
  4522. 966
  4523. 00:50:11,095 --> 00:50:14,272
  4524. when I should only consider
  4525. the safest possible route
  4526.  
  4527. 967
  4528. 00:50:14,315 --> 00:50:16,361
  4529. to achieve success.
  4530.  
  4531. 968
  4532. 00:50:16,404 --> 00:50:19,755
  4533. But I had to stand there for
  4534. 10 minutes agonizing about it.
  4535.  
  4536. 969
  4537. 00:50:19,799 --> 00:50:21,757
  4538. The computer feels nothing.
  4539.  
  4540. 970
  4541. 00:50:21,801 --> 00:50:24,760
  4542. The computer just does
  4543. what it's supposed to do,
  4544.  
  4545. 971
  4546. 00:50:24,804 --> 00:50:26,284
  4547. better and better.
  4548.  
  4549. 972
  4550. 00:50:26,327 --> 00:50:30,288
  4551. ♪♪
  4552.  
  4553. 973
  4554. 00:50:30,331 --> 00:50:32,551
  4555. I want to be AI in this case.
  4556.  
  4557. 974
  4558. 00:50:35,945 --> 00:50:38,861
  4559. But can AI be compassionate?
  4560.  
  4561. 975
  4562. 00:50:38,905 --> 00:50:43,040
  4563. ♪♪
  4564.  
  4565. 976
  4566. 00:50:43,083 --> 00:50:47,827
  4567. I mean, it's everybody's
  4568. question about AI.
  4569.  
  4570. 977
  4571. 00:50:47,870 --> 00:50:51,961
  4572. We are the sole
  4573. embodiment of humanity,
  4574.  
  4575. 978
  4576. 00:50:52,005 --> 00:50:55,269
  4577. and it's a stretch for us
  4578. to accept that a machine
  4579.  
  4580. 979
  4581. 00:50:55,313 --> 00:50:58,794
  4582. can be compassionate
  4583. and loving in that way.
  4584.  
  4585. 980
  4586. 00:50:58,838 --> 00:51:05,105
  4587. ♪♪
  4588.  
  4589. 981
  4590. 00:51:05,149 --> 00:51:07,281
  4591. Part of me
  4592. doesn't believe in magic,
  4593.  
  4594. 982
  4595. 00:51:07,325 --> 00:51:09,805
  4596. but part of me has faith
  4597. that there is something
  4598.  
  4599. 983
  4600. 00:51:09,849 --> 00:51:11,546
  4601. beyond the sum of the parts,
  4602.  
  4603. 984
  4604. 00:51:11,590 --> 00:51:15,637
  4605. that there is at least a oneness
  4606. in our shared ancestry,
  4607.  
  4608. 985
  4609. 00:51:15,681 --> 00:51:20,338
  4610. our shared biology,
  4611. our shared history.
  4612.  
  4613. 986
  4614. 00:51:20,381 --> 00:51:23,210
  4615. Some connection there
  4616. beyond machine.
  4617.  
  4618. 987
  4619. 00:51:23,254 --> 00:51:30,304
  4620. ♪♪
  4621.  
  4622. 988
  4623. 00:51:30,348 --> 00:51:32,567
  4624. So, then, you have
  4625. the other side of that, is,
  4626.  
  4627. 989
  4628. 00:51:32,611 --> 00:51:34,047
  4629. does the computer
  4630. know it's conscious,
  4631.  
  4632. 990
  4633. 00:51:34,091 --> 00:51:37,137
  4634. or can it be conscious,
  4635. or does it care?
  4636.  
  4637. 991
  4638. 00:51:37,181 --> 00:51:40,009
  4639. Does it need to be conscious?
  4640.  
  4641. 992
  4642. 00:51:40,053 --> 00:51:42,011
  4643. Does it need to be aware?
  4644.  
  4645. 993
  4646. 00:51:42,055 --> 00:51:47,365
  4647. ♪♪
  4648.  
  4649. 994
  4650. 00:51:47,408 --> 00:51:52,848
  4651. ♪♪
  4652.  
  4653. 995
  4654. 00:51:52,892 --> 00:51:56,417
  4655. I do not think that a robot
  4656. could ever be conscious.
  4657.  
  4658. 996
  4659. 00:51:56,461 --> 00:51:58,376
  4660. Unless they programmed it
  4661. that way.
  4662.  
  4663. 997
  4664. 00:51:58,419 --> 00:52:00,639
  4665. Conscious? No.
  4666.  
  4667. 998
  4668. 00:52:00,682 --> 00:52:03,163
  4669. No.
  4670. No.
  4671.  
  4672. 999
  4673. 00:52:03,207 --> 00:52:06,035
  4674. I mean, think a robot could be
  4675. programmed to be conscious.
  4676.  
  4677. 1000
  4678. 00:52:06,079 --> 00:52:09,648
  4679. How are they programmed
  4680. to do everything else?
  4681.  
  4682. 1001
  4683. 00:52:09,691 --> 00:52:12,390
  4684. That's another big part
  4685. of artificial intelligence,
  4686.  
  4687. 1002
  4688. 00:52:12,433 --> 00:52:15,741
  4689. is to make them conscious
  4690. and make them feel.
  4691.  
  4692. 1003
  4693. 00:52:17,003 --> 00:52:22,400
  4694. ♪♪
  4695.  
  4696. 1004
  4697. 00:52:22,443 --> 00:52:26,230
  4698. Lipson: Back in 2005, we started
  4699. trying to build machines
  4700.  
  4701. 1005
  4702. 00:52:26,273 --> 00:52:27,709
  4703. with self-awareness.
  4704.  
  4705. 1006
  4706. 00:52:27,753 --> 00:52:33,062
  4707. ♪♪
  4708.  
  4709. 1007
  4710. 00:52:33,106 --> 00:52:37,284
  4711. This robot, to begin with,
  4712. didn't know what it was.
  4713.  
  4714. 1008
  4715. 00:52:37,328 --> 00:52:40,244
  4716. All it knew was that it needed
  4717. to do something like walk.
  4718.  
  4719. 1009
  4720. 00:52:40,287 --> 00:52:44,073
  4721. ♪♪
  4722.  
  4723. 1010
  4724. 00:52:44,117 --> 00:52:45,597
  4725. Through trial and error,
  4726.  
  4727. 1011
  4728. 00:52:45,640 --> 00:52:49,731
  4729. it figured out how to walk
  4730. using its imagination,
  4731.  
  4732. 1012
  4733. 00:52:49,775 --> 00:52:54,040
  4734. and then it walked away.
  4735.  
  4736. 1013
  4737. 00:52:54,083 --> 00:52:56,390
  4738. And then we did
  4739. something very cruel.
  4740.  
  4741. 1014
  4742. 00:52:56,434 --> 00:52:58,653
  4743. We chopped off a leg
  4744. and watched what happened.
  4745.  
  4746. 1015
  4747. 00:52:58,697 --> 00:53:03,005
  4748. ♪♪
  4749.  
  4750. 1016
  4751. 00:53:03,049 --> 00:53:07,749
  4752. At the beginning, it didn't
  4753. quite know what had happened.
  4754.  
  4755. 1017
  4756. 00:53:07,793 --> 00:53:13,233
  4757. But over about a period
  4758. of a day, it then began to limp.
  4759.  
  4760. 1018
  4761. 00:53:13,277 --> 00:53:16,845
  4762. And then, a year ago,
  4763. we were training an AI system
  4764.  
  4765. 1019
  4766. 00:53:16,889 --> 00:53:20,240
  4767. for a live demonstration.
  4768.  
  4769. 1020
  4770. 00:53:20,284 --> 00:53:21,763
  4771. We wanted to show how we wave
  4772.  
  4773. 1021
  4774. 00:53:21,807 --> 00:53:24,113
  4775. all these objects
  4776. in front of the camera
  4777.  
  4778. 1022
  4779. 00:53:24,157 --> 00:53:27,334
  4780. and the AI could
  4781. recognize the objects.
  4782.  
  4783. 1023
  4784. 00:53:27,378 --> 00:53:29,031
  4785. And so, we're preparing
  4786. this demo,
  4787.  
  4788. 1024
  4789. 00:53:29,075 --> 00:53:31,251
  4790. and we had on a side screen
  4791. this ability
  4792.  
  4793. 1025
  4794. 00:53:31,295 --> 00:53:36,778
  4795. to watch what certain
  4796. neurons were responding to.
  4797.  
  4798. 1026
  4799. 00:53:36,822 --> 00:53:39,041
  4800. And suddenly we noticed
  4801. that one of the neurons
  4802.  
  4803. 1027
  4804. 00:53:39,085 --> 00:53:41,087
  4805. was tracking faces.
  4806.  
  4807. 1028
  4808. 00:53:41,130 --> 00:53:45,483
  4809. It was tracking our faces
  4810. as we were moving around.
  4811.  
  4812. 1029
  4813. 00:53:45,526 --> 00:53:48,616
  4814. Now, the spooky thing about this
  4815. is that we never trained
  4816.  
  4817. 1030
  4818. 00:53:48,660 --> 00:53:52,490
  4819. the system
  4820. to recognize human faces,
  4821.  
  4822. 1031
  4823. 00:53:52,533 --> 00:53:55,710
  4824. and yet, somehow,
  4825. it learned to do that.
  4826.  
  4827. 1032
  4828. 00:53:57,973 --> 00:53:59,584
  4829. Even though these robots
  4830. are very simple,
  4831.  
  4832. 1033
  4833. 00:53:59,627 --> 00:54:02,500
  4834. we can see there's
  4835. something else going on there.
  4836.  
  4837. 1034
  4838. 00:54:02,543 --> 00:54:05,851
  4839. It's not just programming.
  4840.  
  4841. 1035
  4842. 00:54:05,894 --> 00:54:08,462
  4843. So, this is just the beginning.
  4844.  
  4845. 1036
  4846. 00:54:10,377 --> 00:54:14,294
  4847. Horvitz: I often think about
  4848. that beach in Kitty Hawk,
  4849.  
  4850. 1037
  4851. 00:54:14,338 --> 00:54:18,255
  4852. the 1903 flight
  4853. by Orville and Wilbur Wright.
  4854.  
  4855. 1038
  4856. 00:54:21,214 --> 00:54:24,348
  4857. It was kind of a canvas plane,
  4858. and it's wood and iron,
  4859.  
  4860. 1039
  4861. 00:54:24,391 --> 00:54:26,828
  4862. and it gets off the ground for,
  4863. what, a minute and 20 seconds,
  4864.  
  4865. 1040
  4866. 00:54:26,872 --> 00:54:29,091
  4867. on this windy day
  4868.  
  4869. 1041
  4870. 00:54:29,135 --> 00:54:31,006
  4871. before touching back down again.
  4872.  
  4873. 1042
  4874. 00:54:33,270 --> 00:54:37,143
  4875. And it was
  4876. just around 65 summers or so
  4877.  
  4878. 1043
  4879. 00:54:37,186 --> 00:54:43,149
  4880. after that moment that you have
  4881. a 747 taking off from JFK...
  4882.  
  4883. 1044
  4884. 00:54:43,192 --> 00:54:50,156
  4885. ♪♪
  4886.  
  4887. 1045
  4888. 00:54:50,199 --> 00:54:51,984
  4889. ...where a major concern
  4890. of someone on the airplane
  4891.  
  4892. 1046
  4893. 00:54:52,027 --> 00:54:55,422
  4894. might be whether or not
  4895. their salt-free diet meal
  4896.  
  4897. 1047
  4898. 00:54:55,466 --> 00:54:56,902
  4899. is gonna be coming to them
  4900. or not.
  4901.  
  4902. 1048
  4903. 00:54:56,945 --> 00:54:58,469
  4904. We have a whole infrastructure,
  4905.  
  4906. 1049
  4907. 00:54:58,512 --> 00:55:01,385
  4908. with travel agents
  4909. and tower control,
  4910.  
  4911. 1050
  4912. 00:55:01,428 --> 00:55:03,778
  4913. and it's all casual,
  4914. and it's all part of the world.
  4915.  
  4916. 1051
  4917. 00:55:03,822 --> 00:55:07,042
  4918. ♪♪
  4919.  
  4920. 1052
  4921. 00:55:07,086 --> 00:55:09,523
  4922. Right now, as far
  4923. as we've come with machines
  4924.  
  4925. 1053
  4926. 00:55:09,567 --> 00:55:12,134
  4927. that think and solve problems,
  4928. we're at Kitty Hawk now.
  4929.  
  4930. 1054
  4931. 00:55:12,178 --> 00:55:13,745
  4932. We're in the wind.
  4933.  
  4934. 1055
  4935. 00:55:13,788 --> 00:55:17,052
  4936. We have our tattered-canvas
  4937. planes up in the air.
  4938.  
  4939. 1056
  4940. 00:55:17,096 --> 00:55:20,882
  4941. ♪♪
  4942.  
  4943. 1057
  4944. 00:55:20,926 --> 00:55:23,885
  4945. But what happens
  4946. in 65 summers or so?
  4947.  
  4948. 1058
  4949. 00:55:23,929 --> 00:55:27,889
  4950. We will have machines
  4951. that are beyond human control.
  4952.  
  4953. 1059
  4954. 00:55:27,933 --> 00:55:30,457
  4955. Should we worry about that?
  4956.  
  4957. 1060
  4958. 00:55:30,501 --> 00:55:32,590
  4959. ♪♪
  4960.  
  4961. 1061
  4962. 00:55:32,633 --> 00:55:34,853
  4963. I'm not sure it's going to help.
  4964.  
  4965. 1062
  4966. 00:55:40,337 --> 00:55:44,036
  4967. Kaplan: Nobody has any idea
  4968. today what it means for a robot
  4969.  
  4970. 1063
  4971. 00:55:44,079 --> 00:55:46,430
  4972. to be conscious.
  4973.  
  4974. 1064
  4975. 00:55:46,473 --> 00:55:48,649
  4976. There is no such thing.
  4977.  
  4978. 1065
  4979. 00:55:48,693 --> 00:55:50,172
  4980. There are a lot of smart people,
  4981.  
  4982. 1066
  4983. 00:55:50,216 --> 00:55:53,088
  4984. and I have a great deal
  4985. of respect for them,
  4986.  
  4987. 1067
  4988. 00:55:53,132 --> 00:55:57,528
  4989. but the truth is, machines
  4990. are natural psychopaths.
  4991.  
  4992. 1068
  4993. 00:55:57,571 --> 00:55:59,225
  4994. Man:
  4995. Fear came back into the market.
  4996.  
  4997. 1069
  4998. 00:55:59,268 --> 00:56:01,706
  4999. Man #2: Went down 800,
  5000. nearly 1,000, in a heartbeat.
  5001.  
  5002. 1070
  5003. 00:56:01,749 --> 00:56:03,360
  5004. I mean,
  5005. it is classic capitulation.
  5006.  
  5007. 1071
  5008. 00:56:03,403 --> 00:56:04,796
  5009. There are some people
  5010. who are proposing
  5011.  
  5012. 1072
  5013. 00:56:04,839 --> 00:56:07,146
  5014. it was some kind
  5015. of fat-finger error.
  5016.  
  5017. 1073
  5018. 00:56:07,189 --> 00:56:09,583
  5019. Take the Flash Crash of 2010.
  5020.  
  5021. 1074
  5022. 00:56:09,627 --> 00:56:13,413
  5023. In a matter of minutes,
  5024. $1 trillion in value
  5025.  
  5026. 1075
  5027. 00:56:13,457 --> 00:56:15,415
  5028. was lost in the stock market.
  5029.  
  5030. 1076
  5031. 00:56:15,459 --> 00:56:18,984
  5032. Woman: The Dow dropped nearly
  5033. 1,000 points in a half-hour.
  5034.  
  5035. 1077
  5036. 00:56:19,027 --> 00:56:22,553
  5037. Kaplan:
  5038. So, what went wrong?
  5039.  
  5040. 1078
  5041. 00:56:22,596 --> 00:56:26,644
  5042. By that point in time,
  5043. more than 60% of all the trades
  5044.  
  5045. 1079
  5046. 00:56:26,687 --> 00:56:29,124
  5047. that took place
  5048. on the stock exchange
  5049.  
  5050. 1080
  5051. 00:56:29,168 --> 00:56:32,693
  5052. were actually being
  5053. initiated by computers.
  5054.  
  5055. 1081
  5056. 00:56:32,737 --> 00:56:34,216
  5057. Man:
  5058. Panic selling on the way down,
  5059.  
  5060. 1082
  5061. 00:56:34,260 --> 00:56:35,783
  5062. and all of a sudden
  5063. it stopped on a dime.
  5064.  
  5065. 1083
  5066. 00:56:35,827 --> 00:56:37,611
  5067. Man #2: This is all happening
  5068. in real time, folks.
  5069.  
  5070. 1084
  5071. 00:56:37,655 --> 00:56:39,526
  5072. Wisz: The short story of what
  5073. happened in the Flash Crash
  5074.  
  5075. 1085
  5076. 00:56:39,570 --> 00:56:42,399
  5077. is that algorithms
  5078. responded to algorithms,
  5079.  
  5080. 1086
  5081. 00:56:42,442 --> 00:56:45,358
  5082. and it compounded upon itself
  5083. over and over and over again
  5084.  
  5085. 1087
  5086. 00:56:45,402 --> 00:56:47,012
  5087. in a matter of minutes.
  5088.  
  5089. 1088
  5090. 00:56:47,055 --> 00:56:50,972
  5091. Man: At one point, the market
  5092. fell as if down a well.
  5093.  
  5094. 1089
  5095. 00:56:51,016 --> 00:56:54,323
  5096. There is no regulatory body
  5097. that can adapt quickly enough
  5098.  
  5099. 1090
  5100. 00:56:54,367 --> 00:56:57,979
  5101. to prevent potentially
  5102. disastrous consequences
  5103.  
  5104. 1091
  5105. 00:56:58,023 --> 00:57:01,243
  5106. of AI operating
  5107. in our financial systems.
  5108.  
  5109. 1092
  5110. 00:57:01,287 --> 00:57:03,898
  5111. They are so prime
  5112. for manipulation.
  5113.  
  5114. 1093
  5115. 00:57:03,942 --> 00:57:05,639
  5116. Let's talk about the speed
  5117. with which
  5118.  
  5119. 1094
  5120. 00:57:05,683 --> 00:57:08,076
  5121. we are watching
  5122. this market deteriorate.
  5123.  
  5124. 1095
  5125. 00:57:08,120 --> 00:57:11,602
  5126. That's the type of AI-run-amuck
  5127. that scares people.
  5128.  
  5129. 1096
  5130. 00:57:11,645 --> 00:57:13,560
  5131. Kaplan:
  5132. When you give them a goal,
  5133.  
  5134. 1097
  5135. 00:57:13,604 --> 00:57:17,825
  5136. they will relentlessly
  5137. pursue that goal.
  5138.  
  5139. 1098
  5140. 00:57:17,869 --> 00:57:20,393
  5141. How many computer programs
  5142. are there like this?
  5143.  
  5144. 1099
  5145. 00:57:20,437 --> 00:57:23,483
  5146. Nobody knows.
  5147.  
  5148. 1100
  5149. 00:57:23,527 --> 00:57:27,444
  5150. Kosinski: One of the fascinating
  5151. aspects about AI in general
  5152.  
  5153. 1101
  5154. 00:57:27,487 --> 00:57:31,970
  5155. is that no one really
  5156. understands how it works.
  5157.  
  5158. 1102
  5159. 00:57:32,013 --> 00:57:36,975
  5160. Even the people who create AI
  5161. don't really fully understand.
  5162.  
  5163. 1103
  5164. 00:57:37,018 --> 00:57:39,804
  5165. Because it has millions
  5166. of elements,
  5167.  
  5168. 1104
  5169. 00:57:39,847 --> 00:57:41,675
  5170. it becomes completely impossible
  5171.  
  5172. 1105
  5173. 00:57:41,719 --> 00:57:45,113
  5174. for a human being
  5175. to understand what's going on.
  5176.  
  5177. 1106
  5178. 00:57:45,157 --> 00:57:52,512
  5179. ♪♪
  5180.  
  5181. 1107
  5182. 00:57:52,556 --> 00:57:56,037
  5183. Grassegger: Microsoft had set up
  5184. this artificial intelligence
  5185.  
  5186. 1108
  5187. 00:57:56,081 --> 00:57:59,127
  5188. called Tay on Twitter,
  5189. which was a chatbot.
  5190.  
  5191. 1109
  5192. 00:58:00,912 --> 00:58:02,696
  5193. They started out in the morning,
  5194.  
  5195. 1110
  5196. 00:58:02,740 --> 00:58:06,526
  5197. and Tay was starting to tweet
  5198. and learning from stuff
  5199.  
  5200. 1111
  5201. 00:58:06,570 --> 00:58:10,835
  5202. that was being sent to him
  5203. from other Twitter people.
  5204.  
  5205. 1112
  5206. 00:58:10,878 --> 00:58:13,272
  5207. Because some people,
  5208. like trolls, attacked him,
  5209.  
  5210. 1113
  5211. 00:58:13,315 --> 00:58:18,582
  5212. within 24 hours, the Microsoft
  5213. bot became a terrible person.
  5214.  
  5215. 1114
  5216. 00:58:18,625 --> 00:58:21,367
  5217. They had to literally
  5218. pull Tay off the Net
  5219.  
  5220. 1115
  5221. 00:58:21,410 --> 00:58:24,718
  5222. because he had turned
  5223. into a monster.
  5224.  
  5225. 1116
  5226. 00:58:24,762 --> 00:58:30,550
  5227. A misanthropic, racist, horrible
  5228. person you'd never want to meet.
  5229.  
  5230. 1117
  5231. 00:58:30,594 --> 00:58:32,857
  5232. And nobody had foreseen this.
  5233.  
  5234. 1118
  5235. 00:58:35,337 --> 00:58:38,602
  5236. The whole idea of AI is that
  5237. we are not telling it exactly
  5238.  
  5239. 1119
  5240. 00:58:38,645 --> 00:58:42,780
  5241. how to achieve a given
  5242. outcome or a goal.
  5243.  
  5244. 1120
  5245. 00:58:42,823 --> 00:58:46,435
  5246. AI develops on its own.
  5247.  
  5248. 1121
  5249. 00:58:46,479 --> 00:58:48,829
  5250. Nolan: We're worried about
  5251. superintelligent AI,
  5252.  
  5253. 1122
  5254. 00:58:48,873 --> 00:58:52,790
  5255. the master chess player
  5256. that will outmaneuver us,
  5257.  
  5258. 1123
  5259. 00:58:52,833 --> 00:58:55,923
  5260. but AI won't have to
  5261. actually be that smart
  5262.  
  5263. 1124
  5264. 00:58:55,967 --> 00:59:00,145
  5265. to have massively disruptive
  5266. effects on human civilization.
  5267.  
  5268. 1125
  5269. 00:59:00,188 --> 00:59:01,886
  5270. We've seen over the last century
  5271.  
  5272. 1126
  5273. 00:59:01,929 --> 00:59:05,150
  5274. it doesn't necessarily take
  5275. a genius to knock history off
  5276.  
  5277. 1127
  5278. 00:59:05,193 --> 00:59:06,804
  5279. in a particular direction,
  5280.  
  5281. 1128
  5282. 00:59:06,847 --> 00:59:09,589
  5283. and it won't take a genius AI
  5284. to do the same thing.
  5285.  
  5286. 1129
  5287. 00:59:09,633 --> 00:59:13,158
  5288. Bogus election news stories
  5289. generated more engagement
  5290.  
  5291. 1130
  5292. 00:59:13,201 --> 00:59:17,075
  5293. on Facebook
  5294. than top real stories.
  5295.  
  5296. 1131
  5297. 00:59:17,118 --> 00:59:21,079
  5298. Facebook really is
  5299. the elephant in the room.
  5300.  
  5301. 1132
  5302. 00:59:21,122 --> 00:59:23,777
  5303. Kosinski:
  5304. AI running Facebook news feed --
  5305.  
  5306. 1133
  5307. 00:59:23,821 --> 00:59:28,347
  5308. The task for AI
  5309. is keeping users engaged,
  5310.  
  5311. 1134
  5312. 00:59:28,390 --> 00:59:29,827
  5313. but no one really understands
  5314.  
  5315. 1135
  5316. 00:59:29,870 --> 00:59:34,832
  5317. exactly how this AI
  5318. is achieving this goal.
  5319.  
  5320. 1136
  5321. 00:59:34,875 --> 00:59:38,792
  5322. Nolan: Facebook is building an
  5323. elegant mirrored wall around us.
  5324.  
  5325. 1137
  5326. 00:59:38,836 --> 00:59:41,665
  5327. A mirror that we can ask,
  5328. "Who's the fairest of them all?"
  5329.  
  5330. 1138
  5331. 00:59:41,708 --> 00:59:45,016
  5332. and it will answer, "You, you,"
  5333. time and again
  5334.  
  5335. 1139
  5336. 00:59:45,059 --> 00:59:48,193
  5337. and slowly begin
  5338. to warp our sense of reality,
  5339.  
  5340. 1140
  5341. 00:59:48,236 --> 00:59:53,502
  5342. warp our sense of politics,
  5343. history, global events,
  5344.  
  5345. 1141
  5346. 00:59:53,546 --> 00:59:57,028
  5347. until determining what's true
  5348. and what's not true,
  5349.  
  5350. 1142
  5351. 00:59:57,071 --> 00:59:58,943
  5352. is virtually impossible.
  5353.  
  5354. 1143
  5355. 01:00:01,032 --> 01:00:03,861
  5356. The problem is that AI
  5357. doesn't understand that.
  5358.  
  5359. 1144
  5360. 01:00:03,904 --> 01:00:08,039
  5361. AI just had a mission --
  5362. maximize user engagement,
  5363.  
  5364. 1145
  5365. 01:00:08,082 --> 01:00:10,041
  5366. and it achieved that.
  5367.  
  5368. 1146
  5369. 01:00:10,084 --> 01:00:13,653
  5370. Nearly 2 billion people
  5371. spend nearly one hour
  5372.  
  5373. 1147
  5374. 01:00:13,697 --> 01:00:17,831
  5375. on average a day
  5376. basically interacting with AI
  5377.  
  5378. 1148
  5379. 01:00:17,875 --> 01:00:21,530
  5380. that is shaping
  5381. their experience.
  5382.  
  5383. 1149
  5384. 01:00:21,574 --> 01:00:24,664
  5385. Even Facebook engineers,
  5386. they don't like fake news.
  5387.  
  5388. 1150
  5389. 01:00:24,708 --> 01:00:26,666
  5390. It's very bad business.
  5391.  
  5392. 1151
  5393. 01:00:26,710 --> 01:00:28,015
  5394. They want to get rid
  5395. of fake news.
  5396.  
  5397. 1152
  5398. 01:00:28,059 --> 01:00:29,974
  5399. It's just very difficult
  5400. to do because,
  5401.  
  5402. 1153
  5403. 01:00:30,017 --> 01:00:32,324
  5404. how do you recognize news
  5405. as fake
  5406.  
  5407. 1154
  5408. 01:00:32,367 --> 01:00:34,456
  5409. if you cannot read
  5410. all of those news personally?
  5411.  
  5412. 1155
  5413. 01:00:34,500 --> 01:00:39,418
  5414. There's so much
  5415. active misinformation
  5416.  
  5417. 1156
  5418. 01:00:39,461 --> 01:00:41,115
  5419. and it's packaged very well,
  5420.  
  5421. 1157
  5422. 01:00:41,159 --> 01:00:44,553
  5423. and it looks the same when
  5424. you see it on a Facebook page
  5425.  
  5426. 1158
  5427. 01:00:44,597 --> 01:00:47,426
  5428. or you turn on your television.
  5429.  
  5430. 1159
  5431. 01:00:47,469 --> 01:00:49,210
  5432. Nolan:
  5433. It's not terribly sophisticated,
  5434.  
  5435. 1160
  5436. 01:00:49,254 --> 01:00:51,691
  5437. but it is terribly powerful.
  5438.  
  5439. 1161
  5440. 01:00:51,735 --> 01:00:54,346
  5441. And what it means is
  5442. that your view of the world,
  5443.  
  5444. 1162
  5445. 01:00:54,389 --> 01:00:56,435
  5446. which, 20 years ago,
  5447. was determined,
  5448.  
  5449. 1163
  5450. 01:00:56,478 --> 01:01:00,004
  5451. if you watched the nightly news,
  5452. by three different networks,
  5453.  
  5454. 1164
  5455. 01:01:00,047 --> 01:01:02,528
  5456. the three anchors who endeavored
  5457. to try to get it right.
  5458.  
  5459. 1165
  5460. 01:01:02,571 --> 01:01:04,225
  5461. Might have had a little bias
  5462. one way or the other,
  5463.  
  5464. 1166
  5465. 01:01:04,269 --> 01:01:05,923
  5466. but, largely speaking,
  5467. we could all agree
  5468.  
  5469. 1167
  5470. 01:01:05,966 --> 01:01:08,273
  5471. on an objective reality.
  5472.  
  5473. 1168
  5474. 01:01:08,316 --> 01:01:10,754
  5475. Well, that objectivity is gone,
  5476.  
  5477. 1169
  5478. 01:01:10,797 --> 01:01:13,757
  5479. and Facebook has
  5480. completely annihilated it.
  5481.  
  5482. 1170
  5483. 01:01:13,800 --> 01:01:17,064
  5484. ♪♪
  5485.  
  5486. 1171
  5487. 01:01:17,108 --> 01:01:19,197
  5488. If most of your understanding
  5489. of how the world works
  5490.  
  5491. 1172
  5492. 01:01:19,240 --> 01:01:20,807
  5493. is derived from Facebook,
  5494.  
  5495. 1173
  5496. 01:01:20,851 --> 01:01:23,418
  5497. facilitated
  5498. by algorithmic software
  5499.  
  5500. 1174
  5501. 01:01:23,462 --> 01:01:27,118
  5502. that tries to show you
  5503. the news you want to see,
  5504.  
  5505. 1175
  5506. 01:01:27,161 --> 01:01:28,815
  5507. that's a terribly
  5508. dangerous thing.
  5509.  
  5510. 1176
  5511. 01:01:28,859 --> 01:01:33,080
  5512. And the idea that we have not
  5513. only set that in motion,
  5514.  
  5515. 1177
  5516. 01:01:33,124 --> 01:01:37,258
  5517. but allowed bad-faith actors
  5518. access to that information...
  5519.  
  5520. 1178
  5521. 01:01:37,302 --> 01:01:39,565
  5522. I mean, this is a recipe
  5523. for disaster.
  5524.  
  5525. 1179
  5526. 01:01:39,608 --> 01:01:43,134
  5527. ♪♪
  5528.  
  5529. 1180
  5530. 01:01:43,177 --> 01:01:45,876
  5531. Urban: I think that there will
  5532. definitely be lots of bad actors
  5533.  
  5534. 1181
  5535. 01:01:45,919 --> 01:01:48,922
  5536. trying to manipulate the world
  5537. with AI.
  5538.  
  5539. 1182
  5540. 01:01:48,966 --> 01:01:52,143
  5541. 2016 was a perfect example
  5542. of an election
  5543.  
  5544. 1183
  5545. 01:01:52,186 --> 01:01:55,015
  5546. where there was lots of AI
  5547. producing lots of fake news
  5548.  
  5549. 1184
  5550. 01:01:55,059 --> 01:01:58,323
  5551. and distributing it
  5552. for a purpose, for a result.
  5553.  
  5554. 1185
  5555. 01:01:59,890 --> 01:02:02,283
  5556. Ladies and gentlemen,
  5557. honorable colleagues...
  5558.  
  5559. 1186
  5560. 01:02:02,327 --> 01:02:04,546
  5561. it's my privilege
  5562. to speak to you today
  5563.  
  5564. 1187
  5565. 01:02:04,590 --> 01:02:07,985
  5566. about the power of big data
  5567. and psychographics
  5568.  
  5569. 1188
  5570. 01:02:08,028 --> 01:02:09,682
  5571. in the electoral process
  5572.  
  5573. 1189
  5574. 01:02:09,726 --> 01:02:12,206
  5575. and, specifically,
  5576. to talk about the work
  5577.  
  5578. 1190
  5579. 01:02:12,250 --> 01:02:14,513
  5580. that we contributed
  5581. to Senator Cruz's
  5582.  
  5583. 1191
  5584. 01:02:14,556 --> 01:02:16,558
  5585. presidential primary campaign.
  5586.  
  5587. 1192
  5588. 01:02:16,602 --> 01:02:19,910
  5589. Nolan: Cambridge Analytica
  5590. emerged quietly as a company
  5591.  
  5592. 1193
  5593. 01:02:19,953 --> 01:02:21,563
  5594. that, according to its own hype,
  5595.  
  5596. 1194
  5597. 01:02:21,607 --> 01:02:26,307
  5598. has the ability to use
  5599. this tremendous amount of data
  5600.  
  5601. 1195
  5602. 01:02:26,351 --> 01:02:30,137
  5603. in order
  5604. to effect societal change.
  5605.  
  5606. 1196
  5607. 01:02:30,181 --> 01:02:33,358
  5608. In 2016, they had
  5609. three major clients.
  5610.  
  5611. 1197
  5612. 01:02:33,401 --> 01:02:34,794
  5613. Ted Cruz was one of them.
  5614.  
  5615. 1198
  5616. 01:02:34,838 --> 01:02:37,884
  5617. It's easy to forget
  5618. that, only 18 months ago,
  5619.  
  5620. 1199
  5621. 01:02:37,928 --> 01:02:41,148
  5622. Senator Cruz was one of
  5623. the less popular candidates
  5624.  
  5625. 1200
  5626. 01:02:41,192 --> 01:02:42,846
  5627. seeking nomination.
  5628.  
  5629. 1201
  5630. 01:02:42,889 --> 01:02:47,241
  5631. So, what was not possible maybe,
  5632. like, 10 or 15 years ago,
  5633.  
  5634. 1202
  5635. 01:02:47,285 --> 01:02:49,374
  5636. was that you can send fake news
  5637.  
  5638. 1203
  5639. 01:02:49,417 --> 01:02:52,420
  5640. to exactly the people
  5641. that you want to send it to.
  5642.  
  5643. 1204
  5644. 01:02:52,464 --> 01:02:56,685
  5645. And then you could actually see
  5646. how he or she reacts on Facebook
  5647.  
  5648. 1205
  5649. 01:02:56,729 --> 01:02:58,905
  5650. and then adjust that information
  5651.  
  5652. 1206
  5653. 01:02:58,949 --> 01:03:01,778
  5654. according to the feedback
  5655. that you got.
  5656.  
  5657. 1207
  5658. 01:03:01,821 --> 01:03:03,257
  5659. So you can start developing
  5660.  
  5661. 1208
  5662. 01:03:03,301 --> 01:03:06,130
  5663. kind of a real-time management
  5664. of a population.
  5665.  
  5666. 1209
  5667. 01:03:06,173 --> 01:03:08,697
  5668. In this case, we've zoned in
  5669.  
  5670. 1210
  5671. 01:03:08,741 --> 01:03:10,699
  5672. on a group
  5673. we've called "Persuasion."
  5674.  
  5675. 1211
  5676. 01:03:10,743 --> 01:03:13,746
  5677. These are people who are
  5678. definitely going to vote,
  5679.  
  5680. 1212
  5681. 01:03:13,790 --> 01:03:16,705
  5682. to caucus, but they need
  5683. moving from the center
  5684.  
  5685. 1213
  5686. 01:03:16,749 --> 01:03:18,490
  5687. a little bit more
  5688. towards the right.
  5689.  
  5690. 1214
  5691. 01:03:18,533 --> 01:03:19,708
  5692. in order to support Cruz.
  5693.  
  5694. 1215
  5695. 01:03:19,752 --> 01:03:22,059
  5696. They need a persuasion message.
  5697.  
  5698. 1216
  5699. 01:03:22,102 --> 01:03:23,800
  5700. "Gun rights," I've selected.
  5701.  
  5702. 1217
  5703. 01:03:23,843 --> 01:03:25,802
  5704. That narrows the field
  5705. slightly more.
  5706.  
  5707. 1218
  5708. 01:03:25,845 --> 01:03:29,066
  5709. And now we know that we need
  5710. a message on gun rights,
  5711.  
  5712. 1219
  5713. 01:03:29,109 --> 01:03:31,111
  5714. it needs to be
  5715. a persuasion message,
  5716.  
  5717. 1220
  5718. 01:03:31,155 --> 01:03:32,591
  5719. and it needs to be nuanced
  5720.  
  5721. 1221
  5722. 01:03:32,634 --> 01:03:34,201
  5723. according to
  5724. the certain personality
  5725.  
  5726. 1222
  5727. 01:03:34,245 --> 01:03:36,029
  5728. that we're interested in.
  5729.  
  5730. 1223
  5731. 01:03:36,073 --> 01:03:39,946
  5732. Through social media, there's an
  5733. infinite amount of information
  5734.  
  5735. 1224
  5736. 01:03:39,990 --> 01:03:42,514
  5737. that you can gather
  5738. about a person.
  5739.  
  5740. 1225
  5741. 01:03:42,557 --> 01:03:45,734
  5742. We have somewhere close
  5743. to 4,000 or 5,000 data points
  5744.  
  5745. 1226
  5746. 01:03:45,778 --> 01:03:48,563
  5747. on every adult
  5748. in the United States.
  5749.  
  5750. 1227
  5751. 01:03:48,607 --> 01:03:51,915
  5752. Grassegger: It's about targeting
  5753. the individual.
  5754.  
  5755. 1228
  5756. 01:03:51,958 --> 01:03:54,352
  5757. It's like a weapon,
  5758. which can be used
  5759.  
  5760. 1229
  5761. 01:03:54,395 --> 01:03:55,962
  5762. in the totally wrong direction.
  5763.  
  5764. 1230
  5765. 01:03:56,006 --> 01:03:58,051
  5766. That's the problem
  5767. with all of this data.
  5768.  
  5769. 1231
  5770. 01:03:58,095 --> 01:04:02,229
  5771. It's almost as if we built the
  5772. bullet before we built the gun.
  5773.  
  5774. 1232
  5775. 01:04:02,273 --> 01:04:04,362
  5776. Ted Cruz employed our data,
  5777.  
  5778. 1233
  5779. 01:04:04,405 --> 01:04:06,407
  5780. our behavioral insights.
  5781.  
  5782. 1234
  5783. 01:04:06,451 --> 01:04:09,541
  5784. He started from a base
  5785. of less than 5%
  5786.  
  5787. 1235
  5788. 01:04:09,584 --> 01:04:15,590
  5789. and had a very slow-and-steady-
  5790. but-firm rise to above 35%,
  5791.  
  5792. 1236
  5793. 01:04:15,634 --> 01:04:17,157
  5794. making him, obviously,
  5795.  
  5796. 1237
  5797. 01:04:17,201 --> 01:04:20,465
  5798. the second most threatening
  5799. contender in the race.
  5800.  
  5801. 1238
  5802. 01:04:20,508 --> 01:04:23,120
  5803. Now, clearly, the Cruz
  5804. campaign is over now,
  5805.  
  5806. 1239
  5807. 01:04:23,163 --> 01:04:24,904
  5808. but what I can tell you
  5809.  
  5810. 1240
  5811. 01:04:24,948 --> 01:04:28,168
  5812. is that of the two candidates
  5813. left in this election,
  5814.  
  5815. 1241
  5816. 01:04:28,212 --> 01:04:30,867
  5817. one of them is using
  5818. these technologies.
  5819.  
  5820. 1242
  5821. 01:04:32,564 --> 01:04:35,959
  5822. I, Donald John Trump,
  5823. do solemnly swear
  5824.  
  5825. 1243
  5826. 01:04:36,002 --> 01:04:38,222
  5827. that I will faithfully execute
  5828.  
  5829. 1244
  5830. 01:04:38,265 --> 01:04:42,226
  5831. the office of President
  5832. of the United States.
  5833.  
  5834. 1245
  5835. 01:04:42,269 --> 01:04:46,273
  5836. ♪♪
  5837.  
  5838. 1246
  5839. 01:04:48,275 --> 01:04:50,234
  5840. Nolan: Elections are
  5841. a marginal exercise.
  5842.  
  5843. 1247
  5844. 01:04:50,277 --> 01:04:53,237
  5845. It doesn't take
  5846. a very sophisticated AI
  5847.  
  5848. 1248
  5849. 01:04:53,280 --> 01:04:57,719
  5850. in order to have
  5851. a disproportionate impact.
  5852.  
  5853. 1249
  5854. 01:04:57,763 --> 01:05:02,550
  5855. Before Trump, Brexit was
  5856. another supposed client.
  5857.  
  5858. 1250
  5859. 01:05:02,594 --> 01:05:04,726
  5860. Well, at 20 minutes to 5:00,
  5861.  
  5862. 1251
  5863. 01:05:04,770 --> 01:05:08,730
  5864. we can now say
  5865. the decision taken in 1975
  5866.  
  5867. 1252
  5868. 01:05:08,774 --> 01:05:10,950
  5869. by this country to join
  5870. the common market
  5871.  
  5872. 1253
  5873. 01:05:10,994 --> 01:05:15,999
  5874. has been reversed by this
  5875. referendum to leave the EU.
  5876.  
  5877. 1254
  5878. 01:05:16,042 --> 01:05:19,828
  5879. Nolan: Cambridge Analytica
  5880. allegedly uses AI
  5881.  
  5882. 1255
  5883. 01:05:19,872 --> 01:05:23,267
  5884. to push through two of
  5885. the most ground-shaking pieces
  5886.  
  5887. 1256
  5888. 01:05:23,310 --> 01:05:27,967
  5889. of political change
  5890. in the last 50 years.
  5891.  
  5892. 1257
  5893. 01:05:28,011 --> 01:05:30,709
  5894. These are epochal events,
  5895. and if we believe the hype,
  5896.  
  5897. 1258
  5898. 01:05:30,752 --> 01:05:33,755
  5899. they are connected directly
  5900. to a piece of software,
  5901.  
  5902. 1259
  5903. 01:05:33,799 --> 01:05:37,194
  5904. essentially, created
  5905. by a professor at Stanford.
  5906.  
  5907. 1260
  5908. 01:05:37,237 --> 01:05:41,415
  5909. ♪♪
  5910.  
  5911. 1261
  5912. 01:05:41,459 --> 01:05:43,635
  5913. Kosinski:
  5914. Back in 2013, I described
  5915.  
  5916. 1262
  5917. 01:05:43,678 --> 01:05:45,593
  5918. that what they are doing
  5919. is possible
  5920.  
  5921. 1263
  5922. 01:05:45,637 --> 01:05:49,293
  5923. and warned against this
  5924. happening in the future.
  5925.  
  5926. 1264
  5927. 01:05:49,336 --> 01:05:51,382
  5928. Grassegger:
  5929. At the time, Michal Kosinski
  5930.  
  5931. 1265
  5932. 01:05:51,425 --> 01:05:52,949
  5933. was a young Polish researcher
  5934.  
  5935. 1266
  5936. 01:05:52,992 --> 01:05:54,994
  5937. working at the
  5938. Psychometrics Centre.
  5939.  
  5940. 1267
  5941. 01:05:55,038 --> 01:06:00,217
  5942. So, what Michal had done was to
  5943. gather the largest-ever data set
  5944.  
  5945. 1268
  5946. 01:06:00,260 --> 01:06:03,481
  5947. of how people
  5948. behave on Facebook.
  5949.  
  5950. 1269
  5951. 01:06:03,524 --> 01:06:07,789
  5952. Kosinski:
  5953. Psychometrics is trying
  5954. to measure psychological traits,
  5955.  
  5956. 1270
  5957. 01:06:07,833 --> 01:06:09,922
  5958. such as personality,
  5959. intelligence,
  5960.  
  5961. 1271
  5962. 01:06:09,966 --> 01:06:11,880
  5963. political views, and so on.
  5964.  
  5965. 1272
  5966. 01:06:11,924 --> 01:06:15,058
  5967. Now, traditionally,
  5968. those traits were measured
  5969.  
  5970. 1273
  5971. 01:06:15,101 --> 01:06:17,712
  5972. using tests and questions.
  5973.  
  5974. 1274
  5975. 01:06:17,756 --> 01:06:19,410
  5976. Nolan: Personality test --
  5977. the most benign thing
  5978.  
  5979. 1275
  5980. 01:06:19,453 --> 01:06:20,715
  5981. you could possibly think of.
  5982.  
  5983. 1276
  5984. 01:06:20,759 --> 01:06:22,065
  5985. Something that doesn't
  5986. necessarily have
  5987.  
  5988. 1277
  5989. 01:06:22,108 --> 01:06:24,197
  5990. a lot of utility, right?
  5991.  
  5992. 1278
  5993. 01:06:24,241 --> 01:06:27,331
  5994. Kosinski: Our idea was that
  5995. instead of tests and questions,
  5996.  
  5997. 1279
  5998. 01:06:27,374 --> 01:06:30,029
  5999. we could simply look at the
  6000. digital footprints of behaviors
  6001.  
  6002. 1280
  6003. 01:06:30,073 --> 01:06:32,553
  6004. that we are all leaving behind
  6005.  
  6006. 1281
  6007. 01:06:32,597 --> 01:06:34,903
  6008. to understand openness,
  6009.  
  6010. 1282
  6011. 01:06:34,947 --> 01:06:37,732
  6012. conscientiousness,
  6013. neuroticism.
  6014.  
  6015. 1283
  6016. 01:06:37,776 --> 01:06:39,560
  6017. Grassegger: You can easily buy
  6018. personal data,
  6019.  
  6020. 1284
  6021. 01:06:39,604 --> 01:06:43,129
  6022. such as where you live, what
  6023. club memberships you've tried,
  6024.  
  6025. 1285
  6026. 01:06:43,173 --> 01:06:45,044
  6027. which gym you go to.
  6028.  
  6029. 1286
  6030. 01:06:45,088 --> 01:06:47,873
  6031. There are actually marketplaces
  6032. for personal data.
  6033.  
  6034. 1287
  6035. 01:06:47,916 --> 01:06:49,918
  6036. Nolan: It turns out, we can
  6037. discover an awful lot
  6038.  
  6039. 1288
  6040. 01:06:49,962 --> 01:06:51,442
  6041. about what you're gonna do
  6042.  
  6043. 1289
  6044. 01:06:51,485 --> 01:06:55,750
  6045. based on a very, very tiny
  6046. set of information.
  6047.  
  6048. 1290
  6049. 01:06:55,794 --> 01:06:58,275
  6050. Kosinski: We are training
  6051. deep-learning networks
  6052.  
  6053. 1291
  6054. 01:06:58,318 --> 01:07:01,278
  6055. to infer intimate traits,
  6056.  
  6057. 1292
  6058. 01:07:01,321 --> 01:07:04,759
  6059. people's political views,
  6060. personality,
  6061.  
  6062. 1293
  6063. 01:07:04,803 --> 01:07:07,806
  6064. intelligence,
  6065. sexual orientation
  6066.  
  6067. 1294
  6068. 01:07:07,849 --> 01:07:10,504
  6069. just from an image
  6070. from someone's face.
  6071.  
  6072. 1295
  6073. 01:07:10,548 --> 01:07:17,033
  6074. ♪♪
  6075.  
  6076. 1296
  6077. 01:07:17,076 --> 01:07:20,645
  6078. Now think about countries which
  6079. are not so free and open-minded.
  6080.  
  6081. 1297
  6082. 01:07:20,688 --> 01:07:23,300
  6083. If you can reveal people's
  6084. religious views
  6085.  
  6086. 1298
  6087. 01:07:23,343 --> 01:07:25,954
  6088. or political views
  6089. or sexual orientation
  6090.  
  6091. 1299
  6092. 01:07:25,998 --> 01:07:28,740
  6093. based on only profile pictures,
  6094.  
  6095. 1300
  6096. 01:07:28,783 --> 01:07:33,310
  6097. this could be literally
  6098. an issue of life and death.
  6099.  
  6100. 1301
  6101. 01:07:33,353 --> 01:07:36,965
  6102. ♪♪
  6103.  
  6104. 1302
  6105. 01:07:37,009 --> 01:07:39,751
  6106. I think there's no going back.
  6107.  
  6108. 1303
  6109. 01:07:42,145 --> 01:07:44,321
  6110. Do you know what
  6111. the Turing test is?
  6112.  
  6113. 1304
  6114. 01:07:44,364 --> 01:07:48,977
  6115. It's when a human interacts
  6116. with a computer,
  6117.  
  6118. 1305
  6119. 01:07:49,021 --> 01:07:50,805
  6120. and if the human doesn't know
  6121. they're interacting
  6122.  
  6123. 1306
  6124. 01:07:50,849 --> 01:07:52,546
  6125. with a computer,
  6126.  
  6127. 1307
  6128. 01:07:52,590 --> 01:07:54,026
  6129. the test is passed.
  6130.  
  6131. 1308
  6132. 01:07:54,070 --> 01:07:57,247
  6133. And over the next few days,
  6134.  
  6135. 1309
  6136. 01:07:57,290 --> 01:07:59,684
  6137. you're gonna be the human
  6138. component in a Turing test.
  6139.  
  6140. 1310
  6141. 01:07:59,727 --> 01:08:02,295
  6142. Holy shit.Yeah, that's right, Caleb.
  6143.  
  6144. 1311
  6145. 01:08:02,339 --> 01:08:04,080
  6146. You got it.
  6147.  
  6148. 1312
  6149. 01:08:04,123 --> 01:08:06,865
  6150. 'Cause if that test
  6151. is passed,
  6152.  
  6153. 1313
  6154. 01:08:06,908 --> 01:08:10,825
  6155. you are dead center of
  6156. the greatest scientific event
  6157.  
  6158. 1314
  6159. 01:08:10,869 --> 01:08:12,958
  6160. in the history of man.
  6161.  
  6162. 1315
  6163. 01:08:13,001 --> 01:08:14,612
  6164. If you've created
  6165. a conscious machine,
  6166.  
  6167. 1316
  6168. 01:08:14,655 --> 01:08:17,615
  6169. it's not the history
  6170. of man--
  6171.  
  6172. 1317
  6173. 01:08:17,658 --> 01:08:19,356
  6174. That's the history
  6175. of gods.
  6176.  
  6177. 1318
  6178. 01:08:19,399 --> 01:08:26,798
  6179. ♪♪
  6180.  
  6181. 1319
  6182. 01:08:26,841 --> 01:08:28,452
  6183. Nolan: It's almost like
  6184. technology is a god
  6185.  
  6186. 1320
  6187. 01:08:28,495 --> 01:08:29,975
  6188. in and of itself.
  6189.  
  6190. 1321
  6191. 01:08:30,018 --> 01:08:33,152
  6192. ♪♪
  6193.  
  6194. 1322
  6195. 01:08:33,196 --> 01:08:35,241
  6196. Like the weather.
  6197. We can't impact it.
  6198.  
  6199. 1323
  6200. 01:08:35,285 --> 01:08:39,593
  6201. We can't slow it down.
  6202. We can't stop it.
  6203.  
  6204. 1324
  6205. 01:08:39,637 --> 01:08:43,249
  6206. We feel powerless.
  6207.  
  6208. 1325
  6209. 01:08:43,293 --> 01:08:44,685
  6210. Kurzweil:
  6211. If we think of God
  6212.  
  6213. 1326
  6214. 01:08:44,729 --> 01:08:46,687
  6215. as an unlimited amount
  6216. of intelligence,
  6217.  
  6218. 1327
  6219. 01:08:46,731 --> 01:08:48,167
  6220. the closest we can get to that
  6221.  
  6222. 1328
  6223. 01:08:48,211 --> 01:08:50,474
  6224. is by evolving
  6225. our own intelligence
  6226.  
  6227. 1329
  6228. 01:08:50,517 --> 01:08:55,566
  6229. by merging with the artificial
  6230. intelligence we're creating.
  6231.  
  6232. 1330
  6233. 01:08:55,609 --> 01:08:58,003
  6234. Musk:
  6235. Today, our computers, phones,
  6236.  
  6237. 1331
  6238. 01:08:58,046 --> 01:09:01,615
  6239. applications give us
  6240. superhuman capability.
  6241.  
  6242. 1332
  6243. 01:09:01,659 --> 01:09:04,662
  6244. So, as the old maxim says,
  6245. if you can't beat 'em, join 'em.
  6246.  
  6247. 1333
  6248. 01:09:06,968 --> 01:09:09,971
  6249. el Kaliouby: It's about
  6250. a human-machine partnership.
  6251.  
  6252. 1334
  6253. 01:09:10,015 --> 01:09:11,669
  6254. I mean, we already see
  6255. how, you know,
  6256.  
  6257. 1335
  6258. 01:09:11,712 --> 01:09:14,933
  6259. our phones, for example, act
  6260. as memory prosthesis, right?
  6261.  
  6262. 1336
  6263. 01:09:14,976 --> 01:09:17,196
  6264. I don't have to remember
  6265. your phone number anymore
  6266.  
  6267. 1337
  6268. 01:09:17,240 --> 01:09:19,198
  6269. 'cause it's on my phone.
  6270.  
  6271. 1338
  6272. 01:09:19,242 --> 01:09:22,070
  6273. It's about machines
  6274. augmenting our human abilities,
  6275.  
  6276. 1339
  6277. 01:09:22,114 --> 01:09:25,248
  6278. as opposed to, like,
  6279. completely displacing them.
  6280.  
  6281. 1340
  6282. 01:09:25,291 --> 01:09:27,380
  6283. Nolan: If you look at all the
  6284. objects that have made the leap
  6285.  
  6286. 1341
  6287. 01:09:27,424 --> 01:09:30,122
  6288. from analog to digital
  6289. over the last 20 years...
  6290.  
  6291. 1342
  6292. 01:09:30,166 --> 01:09:32,080
  6293. it's a lot.
  6294.  
  6295. 1343
  6296. 01:09:32,124 --> 01:09:35,388
  6297. We're the last analog object
  6298. in a digital universe.
  6299.  
  6300. 1344
  6301. 01:09:35,432 --> 01:09:36,911
  6302. And the problem with that,
  6303. of course,
  6304.  
  6305. 1345
  6306. 01:09:36,955 --> 01:09:40,567
  6307. is that the data input/output
  6308. is very limited.
  6309.  
  6310. 1346
  6311. 01:09:40,611 --> 01:09:42,613
  6312. It's this.
  6313. It's these.
  6314.  
  6315. 1347
  6316. 01:09:42,656 --> 01:09:45,355
  6317. Zilis:
  6318. Our eyes are pretty good.
  6319.  
  6320. 1348
  6321. 01:09:45,398 --> 01:09:48,445
  6322. We're able to take in a lot
  6323. of visual information.
  6324.  
  6325. 1349
  6326. 01:09:48,488 --> 01:09:52,536
  6327. But our information output
  6328. is very, very, very low.
  6329.  
  6330. 1350
  6331. 01:09:52,579 --> 01:09:55,669
  6332. The reason this is important --
  6333. If we envision a scenario
  6334.  
  6335. 1351
  6336. 01:09:55,713 --> 01:09:59,543
  6337. where AI's playing a more
  6338. prominent role in societies,
  6339.  
  6340. 1352
  6341. 01:09:59,586 --> 01:10:02,023
  6342. we want good ways to interact
  6343. with this technology
  6344.  
  6345. 1353
  6346. 01:10:02,067 --> 01:10:04,983
  6347. so that it ends up
  6348. augmenting us.
  6349.  
  6350. 1354
  6351. 01:10:05,026 --> 01:10:07,812
  6352. ♪♪
  6353.  
  6354. 1355
  6355. 01:10:07,855 --> 01:10:09,553
  6356. Musk: I think
  6357. it's incredibly important
  6358.  
  6359. 1356
  6360. 01:10:09,596 --> 01:10:12,295
  6361. that AI not be "other."
  6362.  
  6363. 1357
  6364. 01:10:12,338 --> 01:10:14,862
  6365. It must be us.
  6366.  
  6367. 1358
  6368. 01:10:14,906 --> 01:10:18,605
  6369. And I could be wrong
  6370. about what I'm saying.
  6371.  
  6372. 1359
  6373. 01:10:18,649 --> 01:10:20,216
  6374. I'm certainly open to ideas
  6375.  
  6376. 1360
  6377. 01:10:20,259 --> 01:10:23,915
  6378. if anybody can suggest
  6379. a path that's better.
  6380.  
  6381. 1361
  6382. 01:10:23,958 --> 01:10:27,266
  6383. But I think we're gonna really
  6384. have to either merge with AI
  6385.  
  6386. 1362
  6387. 01:10:27,310 --> 01:10:28,963
  6388. or be left behind.
  6389.  
  6390. 1363
  6391. 01:10:29,007 --> 01:10:36,362
  6392. ♪♪
  6393.  
  6394. 1364
  6395. 01:10:36,406 --> 01:10:38,756
  6396. Gourley: It's hard to kind of
  6397. think of unplugging a system
  6398.  
  6399. 1365
  6400. 01:10:38,799 --> 01:10:41,802
  6401. that's distributed
  6402. everywhere on the planet,
  6403.  
  6404. 1366
  6405. 01:10:41,846 --> 01:10:45,806
  6406. that's distributed now
  6407. across the solar system.
  6408.  
  6409. 1367
  6410. 01:10:45,850 --> 01:10:49,375
  6411. You can't just, you know,
  6412. shut that off.
  6413.  
  6414. 1368
  6415. 01:10:49,419 --> 01:10:51,290
  6416. Nolan:
  6417. We've opened Pandora's box.
  6418.  
  6419. 1369
  6420. 01:10:51,334 --> 01:10:55,642
  6421. We've unleashed forces that
  6422. we can't control, we can't stop.
  6423.  
  6424. 1370
  6425. 01:10:55,686 --> 01:10:57,296
  6426. We're in the midst
  6427. of essentially creating
  6428.  
  6429. 1371
  6430. 01:10:57,340 --> 01:10:59,516
  6431. a new life-form on Earth.
  6432.  
  6433. 1372
  6434. 01:10:59,559 --> 01:11:05,826
  6435. ♪♪
  6436.  
  6437. 1373
  6438. 01:11:05,870 --> 01:11:07,611
  6439. Russell:
  6440. We don't know what happens next.
  6441.  
  6442. 1374
  6443. 01:11:07,654 --> 01:11:10,353
  6444. We don't know what shape
  6445. the intellect of a machine
  6446.  
  6447. 1375
  6448. 01:11:10,396 --> 01:11:14,531
  6449. will be when that intellect is
  6450. far beyond human capabilities.
  6451.  
  6452. 1376
  6453. 01:11:14,574 --> 01:11:17,360
  6454. It's just not something
  6455. that's possible.
  6456.  
  6457. 1377
  6458. 01:11:17,403 --> 01:11:24,715
  6459. ♪♪
  6460.  
  6461. 1378
  6462. 01:11:24,758 --> 01:11:26,934
  6463. The least scary future
  6464. I can think of is one
  6465.  
  6466. 1379
  6467. 01:11:26,978 --> 01:11:29,633
  6468. where we have at least
  6469. democratized AI.
  6470.  
  6471. 1380
  6472. 01:11:31,548 --> 01:11:34,159
  6473. Because if one company
  6474. or small group of people
  6475.  
  6476. 1381
  6477. 01:11:34,202 --> 01:11:37,031
  6478. manages to develop godlike
  6479. digital superintelligence,
  6480.  
  6481. 1382
  6482. 01:11:37,075 --> 01:11:40,339
  6483. they can take over the world.
  6484.  
  6485. 1383
  6486. 01:11:40,383 --> 01:11:42,210
  6487. At least when there's
  6488. an evil dictator,
  6489.  
  6490. 1384
  6491. 01:11:42,254 --> 01:11:44,343
  6492. that human is going to die,
  6493.  
  6494. 1385
  6495. 01:11:44,387 --> 01:11:46,998
  6496. but, for an AI,
  6497. there would be no death.
  6498.  
  6499. 1386
  6500. 01:11:47,041 --> 01:11:49,392
  6501. It would live forever.
  6502.  
  6503. 1387
  6504. 01:11:49,435 --> 01:11:51,916
  6505. And then you have
  6506. an immortal dictator
  6507.  
  6508. 1388
  6509. 01:11:51,959 --> 01:11:53,570
  6510. from which we can never escape.
  6511.  
  6512. 1389
  6513. 01:11:53,613 --> 01:12:02,100
  6514. ♪♪
  6515.  
  6516. 1390
  6517. 01:12:02,143 --> 01:12:10,587
  6518. ♪♪
  6519.  
  6520. 1391
  6521. 01:12:10,630 --> 01:12:19,160
  6522. ♪♪
  6523.  
  6524. 1392
  6525. 01:12:19,204 --> 01:12:27,647
  6526. ♪♪
  6527.  
  6528. 1393
  6529. 01:12:27,691 --> 01:12:36,221
  6530. ♪♪
  6531.  
  6532. 1394
  6533. 01:12:36,264 --> 01:12:44,838
  6534. ♪♪
  6535.  
  6536. 1395
  6537. 01:12:51,845 --> 01:12:53,717
  6538. Woman on P.A.:
  6539. Alan. Macchiato.
  6540.  
  6541. 1396
  6542. 01:13:10,951 --> 01:13:17,610
  6543. ♪♪
  6544.  
  6545. 1397
  6546. 01:13:17,654 --> 01:13:24,269
  6547. ♪♪
  6548.  
  6549. 1398
  6550. 01:13:24,312 --> 01:13:30,884
  6551. ♪♪
  6552.  
  6553. 1399
  6554. 01:13:30,928 --> 01:13:37,543
  6555. ♪♪
  6556.  
  6557. 1400
  6558. 01:13:37,587 --> 01:13:44,245
  6559. ♪♪
  6560.  
  6561. 1401
  6562. 01:13:44,289 --> 01:13:50,817
  6563. ♪♪
  6564.  
  6565. 1402
  6566. 01:13:50,861 --> 01:13:57,520
  6567. ♪♪
  6568.  
  6569. 1403
  6570. 01:13:57,563 --> 01:14:04,178
  6571. ♪♪
  6572.  
  6573. 1404
  6574. 01:14:04,222 --> 01:14:10,794
  6575. ♪♪
  6576.  
  6577. 1405
  6578. 01:14:10,837 --> 01:14:17,496
  6579. ♪♪
  6580.  
  6581. 1406
  6582. 01:14:17,540 --> 01:14:24,111
  6583. ♪♪
  6584.  
  6585. 1407
  6586. 01:14:24,155 --> 01:14:30,727
  6587. ♪♪
  6588.  
  6589. 1408
  6590. 01:14:30,770 --> 01:14:37,473
  6591. ♪♪
  6592.  
  6593. 1409
  6594. 01:14:37,516 --> 01:14:44,044
  6595. ♪♪
  6596.  
  6597. 1410
  6598. 01:14:44,088 --> 01:14:47,570
  6599. Woman:
  6600. Hello?
  6601.  
  6602. 1411
  6603. 01:14:47,613 --> 01:14:54,838
  6604. ♪♪
  6605.  
  6606. 1412
  6607. 01:14:54,881 --> 01:15:02,236
  6608. ♪♪
  6609.  
  6610. 1413
  6611. 01:15:02,280 --> 01:15:09,548
  6612. ♪♪
  6613.  
  6614. 1414
  6615. 01:15:09,592 --> 01:15:16,773
  6616. ♪♪
  6617.  
  6618. 1415
  6619. 01:15:16,816 --> 01:15:18,688
  6620. ♪ Yeah, yeah
  6621.  
  6622. 1416
  6623. 01:15:18,731 --> 01:15:20,080
  6624. ♪ Yeah, yeah
  6625.  
  6626. 1417
  6627. 01:15:20,124 --> 01:15:27,261
  6628. ♪♪
  6629.  
  6630. 1418
  6631. 01:15:27,305 --> 01:15:34,442
  6632. ♪♪
  6633.  
  6634. 1419
  6635. 01:15:34,486 --> 01:15:41,580
  6636. ♪♪
  6637.  
  6638. 1420
  6639. 01:15:41,624 --> 01:15:43,234
  6640. ♪ Yeah, yeah
  6641.  
  6642. 1421
  6643. 01:15:43,277 --> 01:15:45,541
  6644. ♪ Yeah, yeah
  6645.  
  6646. 1422
  6647. 01:15:45,584 --> 01:15:51,764
  6648. ♪♪
  6649.  
  6650. 1423
  6651. 01:15:51,808 --> 01:15:57,988
  6652. ♪♪
  6653.  
  6654. 1424
  6655. 01:15:58,031 --> 01:16:04,342
  6656. ♪♪
  6657.  
  6658. 1425
  6659. 01:16:04,385 --> 01:16:10,609
  6660. ♪♪
  6661.  
  6662. 1426
  6663. 01:16:10,653 --> 01:16:13,046
  6664. Hello?
  6665.  
  6666. 1427
  6667. 01:16:13,090 --> 01:16:22,012
  6668. ♪♪
  6669.  
  6670. 1428
  6671. 01:16:22,055 --> 01:16:31,021
  6672. ♪♪
  6673.  
  6674. 1429
  6675. 01:16:31,064 --> 01:16:39,986
  6676. ♪♪
  6677.  
  6678. 1430
  6679. 01:16:40,030 --> 01:16:48,952
  6680. ♪♪
  6681.  
  6682. 1431
  6683. 01:16:48,995 --> 01:16:57,961
  6684. ♪♪
  6685.  
  6686. 1432
  6687. 01:16:58,004 --> 01:17:06,926
  6688. ♪♪
  6689.  
  6690. 1433
  6691. 01:17:06,970 --> 01:17:15,892
  6692. ♪♪
  6693.  
  6694. 1434
  6695. 01:17:15,935 --> 01:17:24,901
  6696. ♪♪
  6697.  
  6698. 1435
  6699. 01:17:24,944 --> 01:17:33,910
  6700. ♪♪
  6701.  
  6702. 1436
  6703. 01:17:33,953 --> 01:17:40,960
  6704. ♪♪
RAW Paste Data
We use cookies for various purposes including analytics. By continuing to use Pastebin, you agree to our use of cookies as described in the Cookies Policy. OK, I Understand
 
Top