upvote
What makes you think that?

Some fields are obviously transformed forever (e.g. math), but we don't have robots walking the streets, and training runs still require gigantic data centers full of advanced chips and a human researcher to wake up and go to the office to kick them off.

reply
There's no shortage of human researches to "kick them off" and won't be in foreseeable future.

Even if datacenter buildout would stop entirely today, there's enough datacenters to continue research. It would be compute constrained, but it won't stop.

reply
There's at least one scenario where I see this realistically unwinding and walking back from the potential for large scale job loss, doom scenarios, etc

If this is the tipping point of a bubble pop, companies that remain won't be able to light stacks of cash on fire and demand 40x multipliers on earnings.

Infrastructure already built would still be there to use, as well as the knowledge of how to build frontier models. Companies coming in would need to build financially viable business models though, almost certainly meaning smaller models with less functionality completing busy work rather than racing to AI.

The bubble pop would be painful either way, but that's a future that at least to me seems realistic meaning it wouldn't yet be a singularity.

reply