"As I mentioned in a previous post, that might mean leaning further into the summary as being a summary of results — describing what’s in the result set, complete with nods to sourcing, rather than trying to turn results answering vastly different questions into a single answer."
FastGPT, bundled with Kagi, does exactly this. My search experience has never been better than it is now. Kagi isn't free, but I'd rather pay in cash than in corrupted search results.
This is such a great example of how an AI/LLM doesn't "think" in the way we understand it! It's just matching words together when it's seen those words go together, and has (to my knowledge, at least) no reliable way of knowing if the resultant output makes any sense. I like your suggestion to have it summarize the results themselves, rather than the content of those results; that seems like something it's much better equipped to do in a way that provides something useful
Could we stop pretending that AI is making the internet better? Its creators are doubling down and saying "if we just tweak it a little" instead of admitting it was a horrible idea from the get-go. We should not be "getting used to" hallucinations, i.e. false information. We should be demanding truth.
"As I mentioned in a previous post, that might mean leaning further into the summary as being a summary of results — describing what’s in the result set, complete with nods to sourcing, rather than trying to turn results answering vastly different questions into a single answer."
FastGPT, bundled with Kagi, does exactly this. My search experience has never been better than it is now. Kagi isn't free, but I'd rather pay in cash than in corrupted search results.
This is such a great example of how an AI/LLM doesn't "think" in the way we understand it! It's just matching words together when it's seen those words go together, and has (to my knowledge, at least) no reliable way of knowing if the resultant output makes any sense. I like your suggestion to have it summarize the results themselves, rather than the content of those results; that seems like something it's much better equipped to do in a way that provides something useful
Could we stop pretending that AI is making the internet better? Its creators are doubling down and saying "if we just tweak it a little" instead of admitting it was a horrible idea from the get-go. We should not be "getting used to" hallucinations, i.e. false information. We should be demanding truth.
Or, just show me the actual results, because I *do* know what I'm looking for.