I assume they all crib from the same training sets, but surely one of the billion dollar companies behind them can make their own?

  • damnthefilibuster@lemmy.worldEnglish
    3·
    2 days ago

    Yeah, they really need to start building RAG supported models. That way they can actually show where they’re getting their data, and even pay the sources fairly. Imagine a RAG or MCP server connecting to Wikipedia, one to encyclopedia.com, and one to stack overflow.