MachuPikacchu on Nostr: The big AI shops have access to a lot of the internet’s content from the era before ...

The big AI shops have access to a lot of the internet’s content from the era before LLM proliferation. This is one motivating factor for why they release models publicly: more AI generated content on the internet creates a moat around their data.

By training and releasing these open source models they create a sort of fog of war for people that come after them since training on AI generated content distorts the output distribution of your model. They’re effectively polluting the digital environment or at least enabling others to do it because it benefits their business model.

Webs of trust are going to become even more important for us as they’ll reduce some of the noise. We need more solutions to this problem though.

#ai #llm