Mitchell_Porter's Shortform

Mitchell_Porter

This is a special post for quick takes by Mitchell_Porter. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.

Has any pause activist proposed a pause-compatible model for the future of frontier AI companies?

I have expressed doubts about the future of the current AI economic model anyway (due to the world economy or just inherent unsustainability), but even if neither scenario happens, for now it seems the future of American AI is being set by the financial imperatives of the companies involved, so it might be desirable to devise a pause scenario within that framework - if that's even possible.

(This hegemony of financial imperatives might change if the Democrats win the midterms and make AI regulation a priority, or if Trump's Super Intelligence task force feels a need to act. And beyond AI for profit, perhaps the final bosses of pause activism are secret military superintelligence projects, and superintelligence projects using harnessed swarms of open-source AI.)

A pause actually could be a short-term financial blessing in disguise. Inference is pretty high margin, and the labs would easily be profitable if they weren't spending on training. I imagine if you take intelligence at the current level and then focus lab resources on commercialization, there's quite a bit of money in it.

That still might whack the companies in the supply chain that are really focused on training compute rather than inference compute (perhaps including Nvidia). But there is a lot of money to be made at the current capability level.

2HowManyAgents7d

I think Roko's AI-safety by ASICS and skeuomorphic AI-safety handles the financial restructuring question very competently.

5StanislavKrym7d

@romeo @Thomas Larsen Could you reevaluate the CapEx forecast from AI-2040?

Mitchell_Porter's Shortform

8

Mitchell_Porter's Shortform

8