5 ms·

I don't get the argument about gpt replacing a search engine. If gpt trains from search engine data how does it get new data? I guess the usage of gpt over time

• postalrat 3y ago
  Why would gpt train from search engine data instead of the actual documents being indexed?
  • ge96 3y ago
    Not just documents, everything, every website that becomes searchable information. How would that work in the future if people don't use search engines, how does the gpt/llm bring that info in. Not saying it's not possible, would there be an API people submit their content into... Edit: actually I wonder if wikipedia will make one (model) they have so much info
    • wfvr 3y ago
      The whole text of wikipedia has been included in training data for basically all LLMs since the start.
      • ge96 3y ago
        That's cool real Hitch Hiker's Guide book
• ranting-moth 3y ago
  Google's search results have been poisoned by SEO. A lot of the time when looking for information about X the search results are dominated either by someone trying to sell me X or they have very superficial information about X just so they can a bit further down try to sell me X. (I'm talking about the unknown X, not Elon's failing project).
  • ge96 3y ago
    I don't disagree with that. I hate trying to get a Towing service and it's a fake website/intercepting the request so they can charge on top/call the actual tow company I was trying to call.