Comment by izzydata

2 years ago

I'd like to see any evidence that suggests AGI is even possible before I care about it wiping out humanity.

11 comments

izzydata

Reply
olalonde  2 years ago

I feel like there's a lot of evidence, for example, the existence of natural general intelligence and the rapidly expanding capacities of modern ANNs. What makes you believe it's not possible? Or what kind of evidence would convince you that it's possible?

  • izzydata  2 years ago

    I believe that it would be possible to make artificial biological intelligence, but that is a whole different can of worms.

    I don't think neural networks, language models, machine learning etc.. are even close to a general intelligence. Maybe there is some way to combine the two. I have seen some demonstrations of very primitive clusters of brain cells being connected to a computer and used to control a small machines direction.

    If there is going to be an AGI I would predict this is how it will happen. While this would be very spectacular and impressive I'm still not worried about it because it would require existing in the physical world and not just some software that can run on any conventional computer.

    • olalonde  2 years ago

      Even if what you say is true (e.g. that the current ANN approach won't lead to AGI), isn't it the case that we can simulate biological cells on computers? Of course, it would push back the AGI timeline by quite a bit, since practically no one is working on this approach right now, but I don't see why it wouldn't be possible in principle.

      4 replies →

brookst  2 years ago

Many of the AGI worriers believe that a fast takeoff will mean the first time we know it's possible will be after the last chance to stop human extinction. I don't buy that myself, but for people who believe that, it's reasonable to want to avoid finding out if it's possible.

VladimirGolovin  2 years ago

You see it every day -- in the mirror. It shows that a kilogram of matter can be arranged into a generally intelligent configuration. Assuming that there's nothing fundamentally special about the physics of the human brain, I see no reason why a functionally similar arrangement cannot be made out of silicon and software.