It's absurd that gen AI is sold as a better search. Like, if you stripped out the "gen" part and used tokenisation to make a better search engine that surfaced original, citable snippets, that could be useful. If you had a way to rank signals that indicated quality answers that could be useful too. Hell, you could even parse queries for intent.
Wait a minute. This is all stuff that search was doing 15 years ago. The "gen" part is just fluff, a vector for introducing significant unreliability.
Absolutely. Exactly what you say.
"Citable snippets", "rank signals", "parse queries for intent". Indeed, these are the elements of meaningful search.
As an academic, few things have pained me more than the fact that most of my colleagues have been drawn into the genAI hype. An intellectual, pedagogical and moral failure of the highest order.
What we should do is develop skills in managing curated knowledge and promote tools that help find and verify sources.
Just make search better. Improve tools, filters, UX. Make it less black box.
Ok, less black box is not what most users want. But the black box has got to be reliable, and I feel we haven't got closer to that in ages, and it's getting worse, not better.
Install Open Science
Get the full app experience
