[tim333]

I have the impression a lot depends on people's past reading and knowledge of what's going on. If you've read the likes of Kurzweil, Moravec, maybe Turing, you're probably going to treat AGI/ASI as inevitable. For people who haven't they just see these chatbots and the like and think those won't change things much.

It's maybe a bit like the early days of covid when the likes of Trump were saying it's nothing, it'll be over by the spring while people who understood virology could see that a bigger thing was on the way.

[morsecodist]

These people's theories (except Turing) are highly speculative predictions about the future. They could be right but they are not analogous to the predictions we get out of epidemiology where we have had a lot of examples to study. What they are doing is not science and it is way more reasonable to doubt them.

[tim333]

The Moravec stuff I'd say is more moderately speculative than highly. All he really said is compute power had tended to double every so long and if that keeps up we'll have human brain equivalent computer in cheap devices in the 2020s. That bit wasn't really a stretch and has largely proved true.

The more unspoken speculative bit is there will then be a large economic incentive for bright researchers and companies to put a lot of effort into sorting the software side. I don't consider LLMs to do the job of general intelligence but there are a lot of people trying to figure it out.

Given we have general intelligence and are the product of ~2GB of DNA, the design can't be that impossible complex, although likely a bit more than gradient descent.