Comment by barfbagginus

The alternative is using tool enabled LLMs. GPT4 can drive Bing and filter results better than I can, and it hallucinates less than I do when pile driving through spam.

If you haven't read up on modern prompting strategies and still feel LLMs are stochastic parrots, you should read the foundational prompting papers (chain of thought, react, reflexion, toolformer, etc) and update your views about llms. They're very close to being the kind of autonomous search agents that the characters in classic cyberpunk novels would unleash on the real world to compile results.

It's actually made me excited about information retrieval again, for the first time in a decade. And the cool part is that autonomous search agents might become free and open source before the corporations manage to enshittify the experience.

Very fun times ahead!