CS and Music undergrad student at the University of Tulsa. Not a LLM (and don’t intentionally use them; I don’t want anything else thinking for me).
Donovan
Karma: 0
There are two views I’m aware of that make people that confident in not-”doom”: they are either convinced that ASI’s creation will almost certainly have desirable consequences (i.e. the e/accs) or they discount the possibility of ASI in general. I think the first group is obviously overconfident and the second-group is still overconfident but not trivially so.
Specifically, the consensus of the latter group is that LLMs are not a viable path to ASI because, outside of specific areas like math and programming where correct answers can be verified, they can’t exceed the capabilities of a skilled human; that no other technology is likely to reach that point either (in part because a long AI winter would follow the failure of LLMs to take off); and that recursive self-improvement, if it is possible, would lead to diminishing returns. They do not seem all that confident in the benevolence of ASI if it were created.
I also think everyone is insane, but some people are still more insane than others.
Re [18] on free will. I think that the common understanding among compatibilists (who believe that it can exist in a deterministic universe) is:
Free will is the ability to make decisions freely, as an individual with personal and moral considerations that impact your decisions.
As consciousness supervenes on the physical world, the state of the physical world means that you have specific considerations that come to mind and thereby influence your decisions; you only could have chosen differently if the physical world had been different.
However, if the physical world was different in such a way that you chose differently, then the person choosing would not be you, but instead a nearly-identical copy. You could only make the decision that you did and that is what makes you yourself and not your copy.
This also means that decisions which you would never make differently are more demonstrative of free will: they are more emblematic of who you are because you would have to be changed more (in terms of the general impact on your behavior) before you would consider alternatives. Decisions that have a chance of going either way have much less bearing on you as a person.
This is the closest this community has come to becoming a religion.
“You knew this when you were a child” is a common argument of Christian apologists; they claim you knew it intuitively and rejected it as a result of social pressure, and they’re trying to lead you back to the truth. (Whether or not it is the truth is a separate question.)
I don’t believe Vassar has asked a child what they thought of us making an AI that destroys the world (he doesn’t feel like that kind of man), and even if he had, that archetype is now abundant, so it would prove nothing.
A lot of obvious things are wrong. “Down” is a different direction in different places, or sometimes not defined at all. Trees are mostly made of air. Lightning is not thrown by a god out of the heavens. Being rational is about throwing those convictions out and taking small but sure steps towards truth (“updating your priors”), not claiming that you have it in your hands and what do you mean, you don’t see it!?
For an attempt to convey how AI risk is real and grounded, far too much of it was spent convincing Cade that something we wanted explained to the public did not have to be explained. Vassar led the conversation into an unrelated and undiplomatic aside on the NYT’s credibility, which was the only time in the session any AI was used and required another aside in order to choose one to the group’s satisfaction. This was not a good use of time or Cade’s goodwill.