How do we make AI a net jobs producer?

[formeruser12]

@Carleas

Specificity limits the disruption, and the generalist AIs will break a lot of things. But it doesn’t seem relevant when thinking about how humans will feel if everything that we value is done better by computers. In the cases where we have some perspective on that question, we find that people feel fine doing something they know a computer could better. I think applies to keeping life interesting more generally: we still value human learning and creativity even where they are significantly surpassed by computers.

You open with speculation, I always try to avoid that when discussing AI, it takes far too much for granted. There’s a difference between a “possibility” and a “certainty”. I don’t consider either AGI or ASI to be certainties, so I don’t factor them in. I think we should be very wary of placing AI on a pedestal above ourselves, and attributing qualities to it that don’t currently exist, and might never exist, it is completely unknown at this stage.

@Carleas

Imaginative output is a function of remixing and randomness.

I think you are vastly oversimplifying that. Our imaginative output is not constrained by previous knowledge, and is not just a “remix” of previous experience. I’ll agree with the “randomness”, but as you well know, computers are not capable of producing randomness, just a facsimile of it, whereas humans are. Many human works can be pointed to that AI just couldn’t have come up with, as it would not have the data in it’s “soup” to do so. We don’t just rehash the old to come up with the completely new, we are capable of making leaps of imagination that AI is clearly not, and we possess inherent qualities to allow that to happen. We don’t even understand fully what those qualities are, so how can they be simulated via AI?

@Carleas

But I don’t think that will be the case so long human interaction with other humans determines reproductive success, and I’m not concerned with the loss of specific learned knowledge or abilities – if we need them again, we’ll learn them again.

It’s the “thinking for you” aspect that I find worrying. AI is a great tool for checking / critiquing logic and reasoning, but it shouldn’t be the originator of those things, and I find when it is left to do such things as tasks, it often fails due to false assumptions. If we can’t solve a problem without “asking the AI”, then we are in big trouble, that’s my impression.

@Carleas

Found this map today: https://www.datacentermap.com/

Thank you, that is very interesting. When I ask about the data-centres, it’s out of legitimate curiosity, it still doesn’t feel like a very long-term or resilient solution, although I agree that under the circumstances it’s more efficient.

@Carleas

This is somewhat subjective, I’ll take a stab at justifying my earlier list:

OK, this is probably to be taken less seriously, it was certainly presented as so, but when I say “markedly improved”, what I mean is in a “quality of life” way. What I’m thinking of is things I could use just fine without pining for the modern equivalents..

@Carleas

This is false.

It’s a fair cop. I haven’t looked at the data, I’ve just always thought there was something “fucky” about the whole thing, and I should probably shut up about it until I do know more. What I said sounds like what someone would say on X, so thanks for the course correction.