Monday, June 20, 2022

Aaronson Switches from Qubits to AI Chatbots

Quantum computer complexity theory Scott Aaronson is leaving the field for a year to join OpenAI. He is jumping from one overhyped field to another.

Aaronson endorsed Google's claims to discovering quantum supremacy, and later quietly backed off. Now he is inspired by a couple of Google employees claiming that Google has invented a sentient chatbot.

I am not sure what the thinking is here. Maybe because Aaronson understands how quantum computers can outdo Turing machines, he will underand how AI will outdo humans? Or vice-versa?

Or because Aaronson has credibly resisted overhyping quantum computers, he will be a credible sage to discuss AI hype?

Previously he announced that Google's chatbot is not really sentient. And said that complexity theorists had taken over the Solvay conference. Some sommenters asked about free will, and he says:

As someone who was actually there, I can tell you that I don’t remember the question of free will ever really coming up at all.
Aomw od rhw commenters seem to believe that studying quantum information theory leads to the conclusion that there can be no free will.

Ben Shapiro, a conservative political activist, comments in a recent short video:

Free will is the single most important principle undergirding any civilization.
Aaronson concedes that when physicists discuss poltitics, they speak as if people have free will.

It appears that the quantum information theory experts do not want to talk about it. To them, free will is both necessary and impossible, and they cannot handle the contradiction.

Update: Here is a newly-posted interview of Sean M. Carroll on free will. Hel says he believes in free will, but only as a term for describing human behavior. He says libertarian free will is absurd, and without any scientific evidence. The Schroedinger equation is deterministic, and makes human choice impossible. But people have an illusion of free will, so it still makes sense to hold them responsible for their choices.

He does not mention Many Worlds theory, but that is why he believes the Schroedinger equation to be deterministic. Maybe he thought that mentioning Many Worlds would undermine his credibility. The textbooks says quantum mechanics predicts randomness, but he believes all things happen in all worlds. Randomness is also an illusion because we do not see the parallel worlds.

Carroll doesn't make any sense. There is evidence for free will every time you make a decision. Quantum mechanics is not deterministic. If free will did not exist, it would not be useful to talk about it.

Aaronson writes:

I have tenure. And I don’t see QC [quantum computing] becoming uninteresting anytime soon (and of course, if it turns out to be impossible for some deep reason, then that will be a revolution in physics). I’m doing this because it’s an opportunity to take a break, learn something new, and possibly make a difference.
Kuhn defined a scientific revolution as a change in viewpoint that has no observable consequences, like changing a reference point in cosmology.

More comments:

“are you … willing to claim that Vladimir Putin is no more responsible for his own outcomes than a tennis ball is responsible for its own outcomes?”

You are saying that Vladimir Putin is not genuinely responsible for starting and continuing the war against Ukraine. So, have the courage of your convictions, and go out and tell your friends and neighbours, and tell the war-crimes tribunals.

I feel like we have to accept the idea that “each and every outcome is 100% due to the laws of nature” for living beings if we are to believe there are laws of nature at all. It seems like the hypothesis that humans, or other sentient beings, can violate the laws of nature through an act of will essentially establishes magic. Most of our work in biological sciences begins with the premise that we can use the scientific method to study the physical processes that combine to produce the behaviors we call “life”, without resorting to magic.

It would be useful if more intellectuals explained their views on free will. It helps in understanding their worldview.

Update: Aaronson did write an 85-page paper on free will in 2013.

1 comment:

  1. If one is going to claim that living things have free will/ responsibility/ creativity, which obviously they DO, as opposed to GPT-3/ Lambda, one has to first model free will.

    The only model of free will, that differentiates free will from non-free will is a model where living things are making inputs to the system. I.e., contrary to your “natura non facit saltus”/ “nature does not make jumps”, we live in a type of world where living things are perceiving their surrounding situation, and taking algorithmic steps in response to this situation, where the “algorithmic steps” are not laws of nature, but just a way of modelling what happens.

    Apart from that, there are no genuine models of free will/ responsibility/ creativity. Any model of free will that claims that free will is a mysterious “subset” of the world that emerges out of complexity, somewhat like a shape that emerges out of cellular automata, not only fails to make explicit what free will is, but fails to differentiate free will from any other aspect of the world.

    Meanwhile, millions of nerds (overwhelmingly male) are getting very excited, and claiming that AIs are no different to people.