Everyone talks about how inference isn’t profitable enough and how, eventually, open-source models will be just as good as frontier models at a fraction of the cost.
But what if, at some point, OpenAI or Anthropic simply stops releasing its best models publicly and develops oracle-level intelligence in-house?
They could then sell access to that intelligence to companies for billions for R&D, or use it themselves for almost anything: finance, drug discovery, engineering, or developing products no one can even conceive of today.
At that point, does it even matter whether inference itself is profitable?
Also what are open source models going to do if they cannot distil the frontier models?
This seems scary since then you truly would have few companies in charge of all new knowledge and innovation generation.
[link] [comments]