Comment

Comment on Cry cry

halvar@lemy.lol ⁨4⁩ ⁨days⁩ ago

Well that’s because in the years leading up to the release of ChatGPT Google became shittier and shittier in performing the service everyone was trying to use it for and everyone needed to work. Then came out chatbots, which basically gave that functionality back. For most people accuracy didn’t matter because they could finally feel again that going to the internet to learn about something is viable.

Then of course AI makes search algortihms even less useful and all the ethical problems about AI are raised as well.

But in the end I can’t really blame people who now use AI instead of Google or something. Finding something with just a search query was always rightfully ours and greedy companies took it to make more ad revenue. Now they are finally giving back a worse version of the original, but one that’s still far superior to what search engines devolved into by now. It’s no surprise most people love and also need that.

source

Sort:hotnew top

jaybone@lemmy.zip ⁨4⁩ ⁨days⁩ ago
Soon with some tweaking the AI answers will serve the shitified shit the SEO driven algorithms shat.

Like: Use Elmer’s TM ©️ glue on your Tombstone Pizza from Kroger to keep your toppings sticky!

So you’ll get ads baked into your wrong answer.

source
Natanox@discuss.tchncs.de ⁨4⁩ ⁨days⁩ ago
There is at least some merit to the technology for this use-case too though (doesn’t even remotely justify the energy costs though, of course). It’s one of the few things LLMs are genuinely good at since it merely requires text ingestion and to regurgitate what was ingested on some way. As long as it’s paired with proper sources (no clue how ChatGPT does it) for all claimed findings it really can be better. Obviously it’s also “better” since it circumvents all the utterly ridiculous trash we usually have to deal with (pop-ups, ads, dark patterns, registration walls, bad search algorithms etc.) which shouldn’t be used as argument.

Paired with the “Thinking” or “Reflection” feature that simulates some basic thinking process (it even enables these things to count the corrrect amount of ‘b’ in ‘blueberry’, wow!) the results are genuinely good (Disclaimer, I only ever tested that with the free tier of Mistral AI - if you really want to use this stuff at least go to them, they’re bound to EU law). I really get why it becomes so popular, and I’d lie if I said I’d never use it myself. Would still prefer if we weren’t going down this cyberpunk timeline though…

source