It’s becoming more and more probable that decentralized AI networks are going to cream Big Tech companies on AI production.

Seems contrarian, I know. But 18 months ago all the AI experts were saying there is nothing to be done about the communication bandwidth bottleneck. That view has now been completely invalidated by Pluralis: we can train on commodity hardware, on consumer devices. GenSyn models, trained by a decentralized RL swarm, is top 3 on HuggingFace. Prime Intellect is heading for massive post-training runs.

It is well known that Big Tech’s proprietary approach slows down innovation. The beauty of web3 is we can move faster with open source research *and* monetize it effectively on decentralized networks.

Big Tech doesn’t completely compete on intelligent models. They are actually competing on dominating and locking in the customer. They do this through good UX and integrations.

But when the deAI stack starts competing at the consumer level as well, customers will have an interesting choice: put my data into this exploitative corporation, or take the privacy preserving (or at least open source) alternative.

If deAI stacks start siphoning off customers, I would be selling Big Tech stock. The big cos will eventually not be able to monetize the progress as fast as open networks.

We have seen web3 networks aggregate a lot of compute. We are going to see them train a large model on that compute soon. But there are also 11 other scaling dimensions on which deAI stacks will compete.

May the best AI models win!