kellogh, 13 days ago

@andrew not worth it. they win either way. you can look at it as waste, or genetic algorithms

JoeStewart, 13 days ago

@andrew @icymi_law @ricmac

#thanksforsharing I wouldn’t otherwise have been aware of either Perplexity or SGE.

I just tried “in massachusetts, what’s the difference between state regulations and state guidelines?”

SGE was no better than search while Perplexity provided a solid answer with citations but the cited URLs did not include “guidelines” - I think it made up its answer.

andrew, 13 days ago

@JoeStewart @icymi_law @ricmac

Just tried the same prompt and got the same results.

I would assume the way these LLMs are working for search is they're answering based on what they "know" and then backsolving from that to find something to cite to.

Right? I would expect actually performing a search and dumping the results of that search into a prompt and running it through the model would be too resource intensive.