Not a member of Pastebin yet?
Sign Up,
it unlocks many cool features!
- -------- Original Message --------
- Subject: What happened to the AlphaGo match in April 2017? -- Re: OGS open sourced, now AlphaGo's turn -- Re: Real potential dangers of Google Deepmind AI gone awry
- Local Time: March 17, 2017 3:39 AM
- UTC Time: March 17, 2017 9:39 AM
- From: TAC777@protonmail.ch
- To: Bo <bc777@protonmail.ch>
- demishassabis@google.com <demishassabis@google.com>, davidsilver@google.com <davidsilver@google.com>, dwyrin@dwyrin.tv <dwyrin@dwyrin.tv>, nick@sibicky.com <nick@sibicky.com>, gohaylee@gmail.com <gohaylee@gmail.com>, elonmuskoffice@teslamotors.com <elonmuskoffice@teslamotors.com>, bowling@cs.ualberta.ca <bowling@cs.ualberta.ca>, hanayeol@gmail.com <hanayeol@gmail.com>, president@usgo.org <president@usgo.org>, bobgilman.aga@gmail.com <bobgilman.aga@gmail.com>, aga@usgo.org <aga@usgo.org>, sebastian@arstechnica.co.uk <sebastian@arstechnica.co.uk>, en@chinanews.com.cn <en@chinanews.com.cn>, remi.coulom@free.fr <remi.coulom@free.fr>, gcp@sjeng.org <gcp@sjeng.org>, david.nolen@gmail.com <david.nolen@gmail.com>, info@samarkand.net <info@samarkand.net>, serial@baduk.or.kr <serial@baduk.or.kr>, secretary@intergofed.org <secretary@intergofed.org>, natalia.kovaleva@eurogofed.org <natalia.kovaleva@eurogofed.org>
- What happened to the AlphaGo match in April 2017 against Ke Jie?
- https://www.reddit.com/r/baduk/comments/5zn8kl/ke_jie_became_the_2nd_player_first_human_to_reach/
- Just like last year, did Google back out at the last minute against its own words and promises yet once again?
- Isn't the Alphabet slogan "Do the right thing"?
- Where is the AlphaGo source code? What happened to doing the right thing? Is it included in that other 99% of Vault7 not yet released?
- Do the right thing. Yeah right.
- https://en.wikipedia.org/wiki/Alphabet_Inc.
- -------- Original Message --------
- Subject: OGS open sourced, now AlphaGo's turn -- Re: Real potential dangers of Google Deepmind AI gone awry -- Re: Open letter to Google and Go Community after the Master/AlphaGo switcharoo - win will AI GO reach human energy parity and still win!?
- Local Time: February 25, 2017 12:02 AM
- UTC Time: February 25, 2017 6:02 AM
- From: bochen7@protonmail.com
- To: Bo <bc777@protonmail.ch>
- demishassabis@google.com <demishassabis@google.com>, davidsilver@google.com <davidsilver@google.com>, dwyrin@dwyrin.tv <dwyrin@dwyrin.tv>, nick@sibicky.com <nick@sibicky.com>, gohaylee@gmail.com <gohaylee@gmail.com>, elonmuskoffice@teslamotors.com <elonmuskoffice@teslamotors.com>, bowling@cs.ualberta.ca <bowling@cs.ualberta.ca>, hanayeol@gmail.com <hanayeol@gmail.com>, president@usgo.org <president@usgo.org>, bobgilman.aga@gmail.com <bobgilman.aga@gmail.com>, aga@usgo.org <aga@usgo.org>, sebastian@arstechnica.co.uk <sebastian@arstechnica.co.uk>, en@chinanews.com.cn <en@chinanews.com.cn>, remi.coulom@free.fr <remi.coulom@free.fr>, gcp@sjeng.org <gcp@sjeng.org>, david.nolen@gmail.com <david.nolen@gmail.com>, info@samarkand.net <info@samarkand.net>, serial@baduk.or.kr <serial@baduk.or.kr>, secretary@intergofed.org <secretary@intergofed.org>, natalia.kovaleva@eurogofed.org <natalia.kovaleva@eurogofed.org>, Frank Chen <frankchen747@gmail.com>, Sue Permantier <suepermantierincentralpoint@gmail.com>, Sarah-Gail Permantier <permantiersarahgail@gmail.com>, enoch_suen@yahoo.com <enoch_suen@yahoo.com>
- This evening OGS did the right thing for the Go community. When is Google going to step up to the plate and man up and do the right thing for the good of the Go community and to set the right examples for the AI world?
- https://forums.online-go.com/t/the-online-go-com-user-interface-is-now-open-source/10987
- 2017 is the year for AI. Google, please live up to the creed of your motto and stop being evil.
- Maybe now that Trump is in charge he can ask NSA to leak the code for the good of all humanity.
- > -------- Original Message --------
- > Subject: Real potential dangers of Google Deepmind AI gone awry -- Re: Open letter to Google and Go Community after the Master/AlphaGo switcharoo - win will AI GO reach human energy parity and still win!?
- > Local Time: January 15, 2017 4:16 PM
- > UTC Time: January 15, 2017 10:16 PM
- > From: bc777@protonmail.ch
- > To: demishassabis@google.com <demishassabis@google.com>, davidsilver@google.com <davidsilver@google.com>, dwyrin@dwyrin.tv <dwyrin@dwyrin.tv>, nick@sibicky.com <nick@sibicky.com>, gohaylee@gmail.com <gohaylee@gmail.com>, elonmuskoffice@teslamotors.com <elonmuskoffice@teslamotors.com>, bowling@cs.ualberta.ca <bowling@cs.ualberta.ca>, hanayeol@gmail.com <hanayeol@gmail.com>
- > bochen2017@protonmail.com <bochen2017@protonmail.com>, president@usgo.org <president@usgo.org>, bobgilman.aga@gmail.com <bobgilman.aga@gmail.com>, aga@usgo.org <aga@usgo.org>, tanguy@le-calve.fr <tanguy@le-calve.fr>, sebastian@arstechnica.co.uk <sebastian@arstechnica.co.uk>, en@chinanews.com.cn <en@chinanews.com.cn>, remi.coulom@free.fr <remi.coulom@free.fr>, gcp@sjeng.org <gcp@sjeng.org>, david.nolen@gmail.com <david.nolen@gmail.com>, info@samarkand.net <info@samarkand.net>, serial@baduk.or.kr <serial@baduk.or.kr>, secretary@intergofed.org <secretary@intergofed.org>, office@intergofed.org <office@intergofed.org>, natalia.kovaleva@eurogofed.org <natalia.kovaleva@eurogofed.org>
- >
- > Nick,
- >
- >
- >
- >
- >
- > It wasn't about boycotting Google but about bringing awareness to the dangers of AI gone awry. Elon Musk himself thought that AI could be more dangerous than nukes and that Google (Deepmind) was of most concern. Narrowed down to the scope of the Go community at large, both in the US at elsewhere in Korea, China, Japan, etc it is clear to see that Google is not doing this (AlphaGo) out of the kindness of its heart. All the Pro's are lapping it up and fighting like dogs over scraps of meat on the bones that Master Google throws their way. They can't see they are being used and thanking AlphaGo for it. Life imitates art no? I hope the irony wasn't lost on you. Google is merely using Go as a platform to advance their AI (NOT the other way around) and a marketing exposure gimmick, nothing more. AI is still in its nascent infancy, we as a humanity can't afford to screw this up and everything Google and Deepmind has shown through its actions causes concern for pause. They are not setting the best of examples for others to follow to say the very least.
- >
- > Some commenters on your video tried to say that it was all the pro's signed NDA's. It wasn't what happened. If Google wanted to do a closed test and made the pro's sign NDA's there is no reason to have the games public at all. Google could have easily done it on their own closed servers if that was the goal, and the larger world would never have known about it unless a pro talked/leaked. And why TWO different fake accounts? (Magic and then Master). Note that at first the games were played under Magic, then when Google saw the win rate was what they had predicted and the outcome favorable they switched over to Master (full on cocky mode) and never went back to using Magic account again. If Google wanted a control account, then Magic should have played as many games as Master, this was far from the case. Reasonable inference could be drawn that Google wanted to test the waters with Magic, when all went well, they started the bragging with Master. Also, notice how although the AI only used 7 seconds per turn, it was also thinking on human time, and it never played any two games at once, meaning very likely they weren't playing this on commodity hardware, otherwise anyone could spin up dozen's of virtual machines on Amazon Web Services and call it a day. AlphaGo is not scaleable down to the single desktop level much less a smartphone and due to the limits of processor physics it never will be.
- >
- > Lee Sedol himself recently remarked that to be actually fair, AG should be given time limits while humans given extra time as opposed to the stone handicap system. The metric used should be energy parity. Human body consumes roughly 100Watts of power, so if AlphaGo is ran on Google's distributed computing datacenter and they can't scale it down to a single machine and still make it viable wins, then they should cap the total thinking time that AlphaGo is allowed to use in order to make it most fair in matches so that AlphaGo uses the same total equivalent power in a match as a human player. Chess has already reached and even surpassed this level, (my smartphone uses less power than a human brain and yet can already use a Chess app on the smartphone to beat all top Chess professionals), but it will be a very long time (if ever) before any GO AI can beat top Go professionals while using the same energy consumption as a human body. The version of AG that beat Lee Sedol used at least 1202 CPUs and 176 GPUs, lets assume a CPU uses only 100Watts and a GPU only 250Watts; the AI was using at least 1,600 times more power than Lee Sedol. Give a top pro 30 minutes per move while limiting AlphaGo to about one second per move, and then it would be a fair game. When Google can win on these terms (without any handicap used) then and only then would it have truly beat humans and solved Go. ("solved" in the same comparative sense that Chess is 'solved' today with Chess apps on phones using less power than human brains yet still beating top Chess pros).
- >
- > Lets see what happens. We live in interesting times indeed. While it will be super fun this year to see AlphaGo give a top pro 4 stones in handicap and still beat humans in a long game, I seriously have my doubts that we will ever see the day that AI can beat humans at Go at energy parity, whether scaling down AlphaGo's energy use to match that of a human, or limiting its time and capping time use to match equivalent human energy use (something we've already achieved for Chess with a smartphone app). But I would love to be proven wrong, unfortunately I really don't think I'm wrong.
- >
- > So tell Andrew Jackson to lighten up. I'm still at 22 Kyu and all the low hanging fruit is still to be had. Anyone offering teaching lessons for a fee?
- >
- > -Bo
- >
- >
- >>> -------- Original Message --------
- >>> Subject: Re: Hey Nick Sibicky -- Re: Open letter to Google and Go Community after the Master/AlphaGo switcharoo - response to Nick's YouTube video 01/14/2017
- >>> Local Time: January 15, 2017 2:40 PM
- >>> UTC Time: January 15, 2017 8:40 PM
- >>> From: nick@sibicky.com
- >>> To: Bo Chen <bochen7@protonmail.com>
- >>>
- >>> Thanks for your thoughts, Bo.
- >>>
- >>> While your calls to boycott Google haven't gone unnoticed, I'm not sure I feel quite the same moral obligation to do so, based on what's transpired with AlphaGo alone.
- >>>
- >>> N
- >
- >
- >
- >
- >
- >
- >> -------- Original Message --------
- >> Subject: Hey Nick Sibicky -- Re: Open letter to Google and Go Community after the Master/AlphaGo switcharoo - response to Nick's YouTube video 01/14/2017
- >> Local Time: January 14, 2017 11:15 PM
- >> UTC Time: January 15, 2017 5:15 AM
- >> From: bochen7@protonmail.com
- >> To: demishassabis@google.com <demishassabis@google.com>, davidsilver@google.com <davidsilver@google.com>, dwyrin@dwyrin.tv <dwyrin@dwyrin.tv>, nick@sibicky.com <nick@sibicky.com>, sundaygolessons@gmail.com <sundaygolessons@gmail.com>, gohaylee@gmail.com <gohaylee@gmail.com>
- >> bochen2017@protonmail.com <bochen2017@protonmail.com>, press@deepmind.com <press@deepmind.com>, president@usgo.org <president@usgo.org>, bobgilman.aga@gmail.com <bobgilman.aga@gmail.com>, communications@usgo.org <communications@usgo.org>, operations@usgo.org <operations@usgo.org>, outreach@usgo.org <outreach@usgo.org>, aga@usgo.org <aga@usgo.org>, bochen7@protonmail.com <bochen7@protonmail.com>
- >>
- >> Nick,
- >>
- >> https://www.youtube.com/watch?v=v8Eh41m7gVA
- >>
- >> Please tell Andrew Jackson to stop crying, so to speak. Anyone can already run Houdini Chess Engine on their computer and already beat all of the world's top Chess pros without exception, but since he works for Google of course he already knows that. No point in regressing to a fully 'solved' game which comprises of 99% tactics. And besides, for the sake of argument, I still think it is much more likely aliens out there are playing Go instead of Chess. Go was never a transcendental game in the sense that it was always brute-force-able at least in theory. The whole "more combinations/configurations/positions than atoms in the universe" part was misleading from the getgo because the visible universe (Hubble Volume) is much smaller than the actual totality of the universe, possibly by trillions of trillions of times, in fact the real universe could be infinite in which case way way way out there is a copy of me beating Master/AlphaGo, food for thought. Did you hear about the new AI that beat the top Poker players and won a lot of money? Now that is something newsworthy.
- >>
- >> Here is what is not going to happen. We are not going to be running any version of AlphaGo on our desktop machines in the near future (five to ten years) even if magically our master the Google overlords were to somehow miraculously open source AlphaGo code tomorrow. (that is also never going to happen, but that is besides the point) Moore's law is dead. Also the problem with dennard scaling. Having more and more processor cores isn't going to help. AlphaGo's AI compute is ran on custom Google ASIC TensorFlow, not exactly general purpose CPU's made by Intel or AMD that we can buy from Amazon or get at Best Buy. And Nvidia CUDA is going no where fast. My guess is Master is three stones stronger than the AlphaGo in March 2016 that beat Lee Sedol, give or take a stone or two. Guess what? On a non-distributed system, meaning on a regular desktop versus Google's datacenter, there is zero change that Master can beat a top pro in a long game. And unless we move away from silicon (not going to happen in our lifetimes), unlike in Chess, there will never be the day that we can install an Go app on our smart phone and have it play better than the top Go players. Ke Jie doesn't have to worry finding a new career anytime soon.
- >>
- >> The guy at your lecture talking about self-flying airplanes being a reality very soon is both right and wrong. Modern airliners essentially already fly themselves, and having been doing so for many decades. In my opinion, Captain Sully was no hero, the Fly By Wire system of the Airbus A320 did all the work, even a five year old kid can use the airbus joystick and glide it onto the water since it is self-stabilizing. The A320 first produced back in 1986, a year after I was born. Boeing's first FBW airliner, the 777, came out back in or around 95, the Windows 95 era. Fully automated airliners will never happen because edge cases can't be easily solved. Boeing's new 777X which won't even come until 2020 added a few more automation features but essentially kept it the same, only thing that really changed on the flightdeck was that now the panels are touchscreen and the CDU to the FMC is now a virtual display. Same for Airbus A350, nothing much changed, infact still using decades old bleed air technology instead of going fully electric like the 787 which has proven to be a complete disaster. And unlike cars, the average service lifespan for an airliner is 27+ years. If Google came out with a self-flying airliner tomorrow and put Boeing, Airbus and COMAC out of business instantly, probably self-flying planes won't become dominant in terms of market-share until your newborn graduates from college and is too old to quality for Obamacare, assuming Trump doesn't keep his word and repeals it soon when he gets into office next week.
- >>
- >> BTW I'm still at 22 kyu, can you give me some Go playing advice? I watched your videos and bought a really expensive Go board and stones but haven't gotten any stronger. I'm beginning to feel like how Lee Sedol felt back in March of last year.
- >>
- >> Sorry I didn't get to met up with you when I was in Seattle. If you are ever in Dallas let me know! I'm still waiting for the day when we can all collect a "basic income" and no one will ever have to work again.
- >>
- >> Bo
- >>
- >>
- >>
- >>
- >>
- >>
- >>
- >>
- >>> -------- Original Message --------
- >>> Subject: Open letter to Google and Go Community after the Master/AlphaGo switcharoo
- >>> Local Time: January 5, 2017 1:24 PM
- >>> UTC Time: January 5, 2017 7:24 PM
- >>> From: bochen2017@protonmail.com
- >>> To: demishassabis@google.com <demishassabis@google.com>, davidsilver@google.com <davidsilver@google.com>, dwyrin@dwyrin.tv <dwyrin@dwyrin.tv>, nick@sibicky.com <nick@sibicky.com>, sundaygolessons@gmail.com <sundaygolessons@gmail.com>, gohaylee@gmail.com <gohaylee@gmail.com>
- >>> bochen2017@protonmail.com <bochen2017@protonmail.com>, press@deepmind.com <press@deepmind.com>, president@usgo.org <president@usgo.org>, bobgilman.aga@gmail.com <bobgilman.aga@gmail.com>, communications@usgo.org <communications@usgo.org>, operations@usgo.org <operations@usgo.org>, outreach@usgo.org <outreach@usgo.org>, aga@usgo.org <aga@usgo.org>
- >>>
- >>> Open letter to Google and Go Community after the Master/AlphaGo switcharoo, lets ask Google for meaningful change for the true betterment and enrichment of all mankind.
- >>>
- >>> We are at crossroads, at a moment in time where we face these moments of truths in the turning point of history in how companies like Google handle A.I. and the manner and ways in which they go about doing it will greatly influence, shape and affect the future events of the world. This is more than just about Go. And we need to ask Google and other innovators to take a much more transparent stand and a more open approach to all of this.
- >>>
- >>> Had Google played fair from the getgo. Eventually A.I. would still have beat humans. But it would have done it in a manner that was fair and square. And had been an accurate truthful representation reflective of the real progress of A.I. at the time. Instead, Google artificially and intentionally, structured the so-called show-casing, of AlphaGo technologies in such an underhanded way using tactics to manipulate and deceive human perception into falsely and incorrectly believing that AlphaGo and Deepmind and Google tech was always better than humans at each and every level and stage of the A.I. advancement and development, and that such was always one step ahead of the curve ahead of the race, so to speak... But as we all know, this was not the case, far from it. And so the public deserved a transparent honest and trustworthy development of AI and not one in which shenanigans are pulled before our eyes for PR stunts and corporate imagery and such.
- >>>
- >>> The recent mystery, mystique and confusion surrounding "master" has now been officially disclosed. This "name change" was most likely motivated by the element of plausible denial-ability. Because only when after Master succeeded in beat the top pros in a series of rapid online games did Google then and only then come clean and to claim the credit. Had the bot failed early on, very likely it would have been pulled, the disclosure and connection never officially made, and Google's AlphaGo still retaining its untarnished brand and records. The using of different names and different accounts goes directly against what Google had earlier promised in its earlier statements in which it had actually went out of its way to affirmatively declare on its own volition that in the future all AlphaGo games would be played under its real name of AlphaGo and AlphaBot. False flag operations (Master had a Korean flag) are usually the specialty reserved for intelligence agencies. So unless Google is taking a page from CIA/NSA, one should never have expected this level of trickery, and there is no justification or excuse.
- >>>
- >>> There is a pattern of this sort of deception and illusion creating trickery going on at Deepmind and Google with respects to AlphaGo. Everything seems like very controlled release and kept too close to the chest, like this sort of artificial illusion of always being perfect and absolutely infallible at every step along the way. Akin to gaming human perception of reality. And not at all academically nor intellectually honest or transparent. For example, had the version of AlphaGo that defeated Fan Hui played Lee Sedol first and right off the bat, there would have been very little chance of AlphaGo securing any victories and very much more probably that Lee Sedol would have won all five games. Had the March 2016 version of AlphaGo played Ke Jie at the time instead of Lee Sedol it is again more likely that Ke Jie would have won many more games. Had AlphaGo and Lee Sedol immediately continued to play five more games after the first five, again it is likely that Lee Sedol would have obtained a higher win rate. Had AlphaGo actuallyed played Ke Jie last year, as previously agreed upon, there was a chance that Ke Jie could have won against it at the point in time. Same for Master. Had Master of right now played in six months or a year later's time any Chinese or Korean bot, Master would also have lost to it. In fact, the reason Google used the new name and kept the identity such a secret was because had Master totally failed to win in the test games, probably its true identity would never have been revealed to the real world.
- >>>
- >>> To Google's credit, what it has proven is that AI Go , when done correctly, will invariably and inevitably far surpass human Go. It appears that is no longer the question and probably hasn't been a point of contention for quite some time. And truthfully, having a machine that plays Go at "God" like levels will help boost all of Go, even the raw enjoyment of human games. So regardless of intent, this push and development has done some good for the Go community and serves to open the door for others to follow...
- >>>
- >>> However, the Google founders claim to be for openness and democracy and for leveling the playing field and not doing evil. If they had spent even a tiny fraction of the time and resources and effort to do whatever it could to in some way open source AlphaGo code (I had contacted Deepmind developers last year, they stated it was possible but would take some work) to be made available for free for the masses at large and not as much on this ridiculous string and incessant series of escalated PR Stunts then humanity itself would be far better off in both the short term and the long run.
- >>>
- >>> I call on Google to do the right thing, to be more transparent where possible, and to in general be a better citizen of the world, more humble, not as proud, and no more of these sleight of hand trickery shenanigan. If Google would be willing to consider open sourcing whatever it could of AG to the community at large, this would be a solid first step in the right direction and on the right side of history.
- >>>
- >>>
- >>> =====
- >>>
- >>>
- >>> So now its proven... Master == AlphaGo.
- >>>
- >>> The name.... many folks thought they named AlphaGo because of Alphabet (Google's parent company, which is also Deepmind's parent company) but rather the truth was it was named "Alpha" in the sense of the Alpha Male... the male that dominates all other males. Now with the name of Master, this not only continues the bragging but also hints of racists (slavery) and brings the symbolic elitism onto another new level not so different from the Nazi Germany regime. (first it was Alpha male, now it is Master race connotations)
- >>>
- >>> We now know for sure that Google is not in the spirit of sportsmanship and indeed with only these five games and nothing else (no source code, no open to play for a fee, etc) it is intellectual misery surpassing the level of what IBM did with DeepBlue when computers first beat Chess players 20 years ago. One would think we have progress into a more open fair society but guess not really.
- >>>
- >>> Especially the name change. Master = AG, then the name change and the mysterious fashion of how this was conducted and etc was entirely unnecessary and unbecoming of a Google that used to vow to "do no evil". (especially since they promised if they ever do play online it will be under the name of AlphaGo, so this bait and switch would make it all double deceptive and hypocritical; at the very least Google is guilty of intentional misrepresentation but although in this case probably not a crime in US/EU/UK, but certainty goes a long way towards being unethical and of bad spirit, almost akin to playing with fake accounts, or having dup accounts)
- >>>
- >>> Google should offer the Go community a full unconditional public apology of the utmost sincerity and generosity.
- >>>
- >>> Previously AlphaGo representations openly denied that another bot was AG, this time around they "no comment"... and since it is so much stronger than anything else out there right now, even Ke Jie stated much stronger than AG back in March, it is a reasonable inference. The fact that right now Master played 50-0, nearly all the top players except for Lee Sodel, (why would he if he already got beat by AG) really lends credence to the theory that Master is Google related and does not bode well for the future of Go due to Google's underhanded tactics and shenanigans.
- >>>
- >>> I say boycott Google.
- >>>
- >>> Edit: We had so many naysayers on here saying that (wrongfully predicting) there was no way Master was AlphaGo etc..... Seems like Google intentionally fooled a most part of the Go community and should be utterly ashamed of its actions, in direct contradiction to what it earlier had affirmatively promised on its own volition that it would play under "AlphaGo" and "AlphaBot" and instead it did the "Master" bait and switch...
- >>>
- >>>
- >>>
- >>>
- >>
Add Comment
Please, Sign In to add comment