> But where is your startup going to find millions of users for extra training data?

The (big) community. There are existing decentralized approaches to accomplish this. And I am not mentioning the data that could be consumed beyond DMCAs.

The startups will use the existing LLMs to bootstrap theirs, getting access to the data by proxy. This will transfer cash to the early LLMs but give a means for startups to get going.

Some startups will use the existing LLMs to bootstrap theirs, getting access to the data by proxy. This will transfer cash to the early LLMs but give a means for startups to get going.

everywhere? what moat does openAI have?

Microsoft Office suite