Chef Winters [$BTC]

I’m often asked about the possibility of using frontier models in an open source environment. I believe that one way or another, frontier companies will eventually adopt the same approach as @near_ai has to privacy, secure enclaves, and end-to-end encryption. It’s possible to encrypt even a closed model and upload it to a decentralized cloud like NEAR AI Cloud. The weights are guaranteed to be hidden from everyone, but inference can be monetized, while users are guaranteed that their data won’t be visible to the model owner. Right now, the frontier model companies are training on user data, but that window is probably closing and most training will be done on synthetic data. So they won't be as incentivized to keep or utilize users' data. Not to mention, over time user data will become even more of a liability than it is already. Enterprises are already paying the big AI companies not to train on their data, albeit with verbal trust rather than mathematical guarantees, but it’s clearly possible. So NEAR isn't working against the frontier model builders. They can use our infra or our approach. Everyone can – and should – have a more User-Owned AI paradigm where frontier models are verifiable and it will be better for everyone.

@Censorship0x +$1,255 from claims -$301 in activation + boosts 150 mining sessions, 13 paid out. Stopped 50 swaps. Stopped Another satisfied Pond user. x.com/genxdegenerate…