The open-source AI boom is built on Big Tech’s handouts. How long will it last?

Big Tech is built on open-source handouts.

And open-source being built on Big Tech's research and models doesn't mean it relies on them. See: the many papers published by Standford and co. which rival OpenAI's research, RedPajama (https://www.together.xyz/blog/redpajama) which rivals LLaMa. And these are built with open alternatives, there'd be more motivation without.

Big Tech has the advantage of compute and resources, but it's probable that open-source has the advantage of research. Since companies may have a few very talented individuals, but open-source has way more people and includes academia.

These tools are the result of academic work, funded by the government for decades, and by user content, also generated for decades.

Corporate AI is just 90% selling back public goods at a markup. They should be legally forced to open source.

The data that the models are built from comes from the public. It'll be a long fight if Big Tech decides that the public's data, however referenced in whatever model, is no longer the public's.

They're fully welcome to keep proprietary the data that is already proprietary. But so far, that's not been their source.

From the horses mouth https://www.semianalysis.com/p/google-we-have-no-moat-and-ne...