←back to thread

207 points lexandstuff | 1 comments | | HN request time: 0.259s | source
Show context
ActorNightly ◴[] No.44477104[source]
If you are going to write anything about AGI, you should really prove that its actually possible in the first place, because that question is not really something that has a definite yes.
replies(3): >>44477147 #>>44477223 #>>44477246 #
mitthrowaway2 ◴[] No.44477246[source]
For most of us non-dualists, the human brain is an existence proof. Doesn't mean transformers and LLMs are the right implementation, but it's not really a question of proving it's possible when it's clearly supported by the fundamental operations available in the universe. So it's okay to skip to the part of the conversation you want to write about.
replies(3): >>44477386 #>>44477553 #>>44477701 #
1. dinkumthinkum ◴[] No.44477553[source]
The human brain is an existence proof? I think that phrase doesn’t mean what you think it means. I don’t think dualist or non-dualist means what you think it means either. When people are talking about AGI, they are clearly talking about something the human research community is actually working towards. Therefore, they are talking about computing equivalent to a Turing machine and using using hardware architecture very similar to what has been currently conceived and developed. Do you have any evidence that the human brains works in such a way? Do you really think that you think and solve problems in that way? Consider simple physics. How much energy is needed and heat produced to train and run these models to solve simple problems. How much of the same is needed and produced when you would solve a sheet of calculus problems, solve a riddle, or write a non-trivial program? Couldn’t you realistically do those things with minimal food and water for a week, if needed? Does it actually seem like the human brain is really at all like these things and is not fundamentally different? I think this is even more naive than if you had proposed “Life exists in the universe, so of course we can create it in a lab by mixing a few solutions.” I think the latter is far likelier and conceivable and even that is still quite an open question.