38C3 - Waiter, There's An LLM In My Search!
ฝัง
- เผยแพร่เมื่อ 7 ก.พ. 2025
- media.ccc.de/v...
This year Large Language Models (LLMs) in search engines told us to put glue on our pizza and eat a small rock every day. This is not ideal, and the consequences of "AI Overviews" and similar features could even be deadly for some people, like mushroom foragers. Maybe it's time for a new sort of search? In this talk I'll sketch out some possible futures and look at how we can put search back in the hands of the searcher. Also, there will be memes!
Overall, the state of search right now is: not good. Search engine results are full of AI generated sludge, SEO spam and self-dealing by providers. This talk will look at the options that are open to us to improve search somewhat, including a few tips and tricks that anyone can take advantage of today to make hyperscale search providers like Google more functional again. But in many ways the most interesting question is whether we can find ways to discover stuff online that don't rely on a handful of hyperscale providers to do all the web crawling and indexing, and servicing of people's queries. In particular, what would happen if search was federated - how could we make that scaleable and performant, and what can we learn from the fediverse?
martinh
events.ccc.de/...
#38c3
Licensed to the public under creativecommons...
I wonder if we’ve reached the point where commercial search engines are so degraded due to monetization that we need to consider legally declaring them to be public utilities and heavily regulating them. Competition isn’t working. Remember when Google search was really good, and always found what you wanted if it existed online? We need to get back to that point.
I highly agree. Being old enough to remember my (successful) swap from Alta Vista to Google, I’d wonder how many younger than say 40 who can _imagine_ public ownership?
I mean, a certain majority need to _want_ such a thing to happen, to possibly _make_ it happen. 😊
this is why I started paying for a search engine.
@@JmbFountain Oh?
Wow got to add all these features great talk
My immediate reaction to the idea of domain specific searches is rejection.
For starters, who gets to define what categories there are, and what belongs to which categories?
Also, in a world of surveillance capitalism, such a search engine would basically be serving up user profiles on a silver platter by telling them what categories/domains someone is interested in.
what stops engines from handing over every query users make already?
great talk thanks :)