Do you have a dumb question that you're kind of embarrassed to ask in the main thread? Is there something you're just not sure about?
This is your opportunity to ask questions. No question too simple or too silly.
Culture war topics are accepted, and proposals for a better intro post are appreciated.
Jump in the discussion.
No email address required.
Notes -
Is it not inevitable that these 'frontier AI labs' calling for regulation are going to get eaten alive by the half dozen big tech companies that have accountants overseeing the nerds? The business case will prevent AI risk, the market won't stay solvent long enough to present X risk.
I may not be educated, but I can sanity check. This stuff is really expensive and when the bill comes due someone who has the revenue to pay it gets the assets.
They are operating on the idea that one of the frontier companies will make a major breakthrough that creates an insurmountable lead. Using customer data and a superhuman AGI they'll be able to immediately move into multiple high profit industries and take over.
The problem they are having is that so far training better models seems to be able to suck up an unlimited amount of money. That was unexpected.
Also the open models have been better at keeping up than they expected. You can fine tune the open models for specific tasks and get "good enough" results.
So they are in a weird spot since there are so man unknown unknowns. There are potentially massive returns. The huge costs are all in training models, so if they cut that back and just run inference with existing models then they can immediately become profitable. They can get to work doing low level speed optimizations on existing models and build out more capacity that way.
Right now they're going for a moon shot, but they can scale back training without being forced to sell of the companies at the risk of losing their leads.
Don't they need to service their debt as well?
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link