Some researchers believe that the creation of a “theory of everything” is quite possible, but humans will not be able to handle it, unlike super-smart AI.
Physics is known to study the mysteries of nature. However, the deeper people delved into studying them, the more questions they had. The main question was formulated by Stephen Hawking: Is there one common explanation for all phenomena? Hawking believed that “the theory of everything” would be the answer to this fundamental question. The scientist also believed that deriving a theory of everything was realistic. And some of today’s researchers support him, believing, however, that it would be within the power of a superpowered computer, not a human being. The science editor of nytimes.com, Dennis Overbay, discusses this possibility with a number of physicists and computer scientists in his column.
Why researchers are teaching AI to “think like a physicist.”
Overbay reports that at least one research institute is seriously engaged in “pumping up” computational systems in physics. That’s the Institute for Artificial Intelligence and Fundamental Interactions, opened by MIT professor Max Tegmark and led by physicist Jesse Thaler. According to the latter, the goal is for computers to learn to “think like physicists. And the scientists have already succeeded in doing something – in 2020, they injected data from the famous Feynman Lectures in Physics textbook into the AI system, and eventually it was able to analyze it for patterns and reconstruct all 100 formulas that were considered in the book.
Dr. Tegmark believes that it would not be difficult for a neural network to detect new particles based on data from the Large Hadron Collider. However, the researcher acknowledges that there are limits to what machines can do that are not capable of “figuring out” things like Heisenberg’s uncertainty principle or relativity theory on their own.
“The reason I work on improving AI systems is to be able to understand the principle of how a superpowered neural network works, if we manage to create one. Otherwise, I fear the consequences: the idea that we simply won’t understand how super-intelligent AI works scares me,” says Tegmark.
Dr. Thaler, claims that he has been able to “encode” some of his knowledge into the machine, which will then provide answers that are easier for him to interpret.
“It’s like a dialogue between a man and a machine. The computer gradually stops being just a box that makes decisions for you and that you don’t understand,” he says.
Recently, Dr. Thaler and his colleagues loaded a large array of data from the Large Hadron Collider into a neural network, and without any additional programs, it was able to distinguish between quarks and gluons (types of elementary particles – ed.), not even knowing what they are. Then the machine was asked the question: what are quarks? And it found the right answer – (so-called) up-quarks and down-quarks.
“The neural network, having no idea about quantum field theory, managed to find the patterns. I was shocked,” the researcher shares.
Will AI be able to create a theory of everything?
“Some physicists believe that the next big leap in technological development will come with the emergence of artificial intelligence for quantum computers. Such machines are still in development, but have great promise,” the author writes, citing Seth Lloyd, a quantum computing expert at MTI, and asks
Could a machine formulate the principles of quantum field theory or relativity theory?
Could it lead to a theory that we humans cannot understand?
Could we end up in a “matrix” or in a world controlled by SkyNet, like in the movie “Terminator”?
All these questions Dennis Overbye asked physicists and computer scientists. And here’s what they answered.
Jaron Lanier, a computer engineer at Microsoft, believes that predictions of an “almighty AI” are an exaggeration. – is an exaggeration.
“An AI system is not a creature like a cat. It’s just a working algorithm,” he said.
Steven Weinberg, a Nobel Prize-winning professor at the University of Texas at Austin, called the suggestion that humans may not be smart enough to understand a theory of everything that can be formulated by a machine “disturbing.”
“But I suspect that we also won’t be smart enough to design a computer that can create that theory,” the scientist responded.