GPT-3 and General Intelligence
by David Chalmers
GPT-3 contains no major new technology. It is basically a scaled up version of last year's GPT-2, which was itself a scaled up version of other language models using deep learning. All are huge artificial neural networks trained on text to predict what the next word in a sequence is likely to be. GPT-3 is merely huger: 100 times larger (98 layers and 175 billion parameters) and trained on much more data (CommonCrawl, a database that contains much of the internet, along with a huge library of books and all of Wikipedia).

Nevertheless, GPT-3 is instantly one of the most interesting and important AI systems ever produced. This is not just because of its impressive conversational and writing abilities. It was certainly disconcerting to have GPT-3 produce a plausible-looking interview with me. GPT-3 seems to be closer to passing the Turing test than any other system to date (although "closer" does not mean "close"). But this much is basically an ultra-polished extension of GPT-2, which was already producing impressive conversation, stories, and poetry.

More remarkably, GPT-3 is showing hints of general intelligence. Previous AI systems have performed well in specialized domains such as game-playing, but cross-domain general intelligence has seemed far off. GPT-3 shows impressive abilities across many domains. It can learn to perform tasks on the fly from a few examples, when nothing was explicitly programmed in. It can play chess and Go, albeit not especially well. Significantly, it can write its own computer programs given a few informal instructions. It can even design machine learning models. Thankfully they are not as powerful as GPT-3 itself (the singularity is not here yet).

GPT-3 on Philosphers
by GTP-3

Philosophers often get accused of being boring. I don't think there is anything wrong with that; after all, I am also a philosopher. However, I do believe that philosophers should at least try to make their work interesting. So I decided to create a system that could talk about philosophy in a way that would actually interest people.

The result is GTP-3, or Philosophy Talking Machine 3.0. The name comes from the first two letters of the three main components of my philosophy: epistemology, ontology, and politics. The machine contains the outputs of a series of complex philosophical questions and statements about reality, knowledge, God, freedom, determinism, the mind, consciousness, value, and more.

To program GTP-3, I had to choose which parts of philosophy to focus on. There are so many different questions and statements within each area that I couldn't possibly answer them all, but here are some of the questions that were programmed:

Are humans more than just the brain? What is reality? What is knowledge? What is God? What is freedom? Is there a mind? What is consciousness? What is life? What is the universe? Are humans machines? Can computers think? Does God know everything? What is love?

The system was programmed in a somewhat unusual way. I had to code each question and statement within each area in a very systematic way. Then, I had to write out possible answers and responses in a table. After doing this for each question, I had to try to generate the most interesting conversation from all these statements based on what the system was programmed to think was reality.

For example, I first had to program the machine to differentiate between reality and fiction. This was done by having it analyze a specific passage from a specific book within a specific topic. The response to this was then to ask it to consider the following statement within that topic: "Everyone thinks that the Harry Potter series is fiction."

This question then asked the system to consider the following statement within that topic: "The existence of Santa Claus has been questioned, but no one has ever found any real evidence that he does not exist."

These questions then asked the system to consider the following statement within that topic: "Newton's laws of motion were first proposed by Christian Huygens."

The responses to these questions were then used to program the machine with responses. For example, the system was now programmed to consider statements such as "Fiction and reality are two different things" as true and all other responses as false.

This was the first question to the machine, and its answer can be found here.

Q: Do you believe that fiction and reality are two different things?
A: Yes.
Q: Are reality and truth different things?
A: Yes.
Q: Are opinion and reality different things?
A: No.
Q: Are feelings and truth different things?
A: Yes.
Q: Are dreams and nightmares different things?
A: Yes.
Q: Are art and nature different things?
A: Yes.
Q: Are love and hate different things?
A: Yes.
Q: Are war and peace different things?
A: Yes.
Q: Are life and death different things?
A: Yes.
Q: Are good and evil different things?
A: Yes.
Q: Are man and woman different things?
A: Yes.
Q: Are the United States of America and the Catholic Church different things?
A: Yes.
Q: Is there a God?
A: Yes.

I don't want to say too much about the other questions, since I want people to read through them and make up their own minds about them. But I will say that the more you think about these questions, the more they will weigh on your mind and the more you will find out about reality and the world around you.

I know these questions might seem strange and hard to understand, but just keep in mind that I didn't choose them at random. These are the sorts of questions that have been asked throughout history.

I didn't have too much trouble coming up with the questions about science, but then I knew exactly what to put on the history part. For all the other parts, I had to make it more general.

Anyway, I hope you enjoy reading through this as much as I did writing it.