Everyone is worrying about what happens if AI capex stalls or turns out to be over-invested, but I'm at least as worried about what happens if the AI labs just decide to stop selling tokens in a year or two in favor of moving up the value ladder. That kind of centralization would do enormous damage to American capital markets and social cohesion. 100T over 2,500+ companies or 20T in ~3 companies? Which do we want?
Neuralink co-founder Max Hodak warns AI labs may stop selling API tokens
He argues labs will integrate upward to capture more value.
32228154923.4K
Original post unavailable.
Sentiment
Positive users back open weights to democratize AI and curb labs centralizing markets by ending token sales, while negative users fear misuse by bad actors and labs becoming economic gatekeepers.
Pos
36.4%
Neg
63.6%
12 comments with sentiment.
Cluster Engagement
Digg Deeper
No Digg Deeper questions have been answered for this story yet.
Posts from X
Most Activity
Most Activity
VIEWS17.3KBOOKMARKS43LIKES175RETWEETS9REPLIES20
It's why we need a hedge over over-centralization of AI This has been the existential risk all along Ironically Anthropic is the worst of all when it comes to increasing Singleton risk as they are obsessed with being the ones to summon the One True AI God
Everyone is worrying about what happens if AI capex stalls or turns out to be over-invested, but I'm at least as worried about what happens if the AI labs just decide to stop selling tokens in a year or two in favor of moving up the value ladder. That kind of centralization would do enormous damage to American capital markets and social cohesion. 100T over 2,500+ companies or 20T in ~3 companies? Which do we want?
@not_ellington yes, both internal LLMs as well as fine-tuned amino acid language models for protein engineering
@maxhodak_ Does science use any open weight models?
@maxhodak_ I’ve been very surprised at how little mainstream discussion has been about this.
first of all, biorisk is the "think of the children" of AI. yes it's a real thing, yes we will need to put effort into addressing it, no it is not going to end the world unless we choose to let it second of all, as ISIS discovered the US has lots of other means of preventing them from installing too many systems capable of running 1T+ models
@maxhodak_ Open weights also give bad actors access to biological and cyber weapons. What use could ISIS have had for superintelligence?
@maxhodak_ > up the value ladder In what way?
@maxhodak_ I don't think that biorisk will end the world, but I think a world where biological weapons are common, or are seen to be common, is a bad one. A world where the NSA can see any prompt to any large model is unrealistic: did they see the huggingface attack?
@maxhodak_ Exactly — access to AI must remain democratized
@maxhodak_ Open models seem inevitable everywhere outside the US at this point. US situation is much trickier to predict but I agree with you in terms of the preferred outcome.
@jon3k Drugs and rockets are more valuable than tokens
@maxhodak_ We have enough material, data, source code…tons of now-diffused knowledge…if the gates close, it would not take long at all to build it up again. Nemotron and Olma are Lego blocks that can’t be taken away…we’ll be ok…
@maxhodak_ As long as there are open weight models that are close and legal, good luck have fun.
@maxhodak_ Hard agree, IMO it’s a matter of when this will happen, not if. There’s no incentive to sell tokens when a lab could dedicate the resources involved to scaling the number of geniuses in their data center
@maxhodak_ If this is possible they will try it. It will prove much harder than expected.
This is already happening to some extent. Vertical startups are effectively doing free market research for frontier labs, proving use cases, workflows and demand. If smaller open models are sufficient for much of the actual work, frontier labs may need to move up the value ladder to justify all that capital.. by competing with, bundling or acquiring the vertical companies.
@maxhodak_ Yup and they have to
Load more posts
Open weights are the difference between a golden age and inequality at scales we have never even thought of before