• makeasnek@lemmy.ml
    link
    fedilink
    English
    arrow-up
    2
    ·
    23 days ago

    We need decentralized, federated search. I remember YaCy from years ago was attempting this. Anybody know if there’s anybody actively working on this?

        • AJ Sadauskas@aus.social
          link
          fedilink
          arrow-up
          2
          ·
          23 days ago

          @makeasnek On a broader note, I think possibly the best approach for decentralised, open-sourced web search might be an evolution on the SearXNG model.

          At the top of the funnel, you have meta search engines that query and aggregate results from a number of smaller niche search engines.

          The metasearch engines are open source, anyone with a spare server or a web hosting account can spin one up.

          For some larger sites that are trustworthy, such as Wikipedia, the site’s own search engine might be what’s queried.

          For the Fediverse and other similar federated networks, the query is fed through a trusted node on the network.

          And then there’s a host of smaller niche search engines, which only crawl and index pages on a small number of websites vetted and curated by a human.

          (Perhaps on a particular topic? Or a local library or university might curate a list of notable local websites?)

          (Alternatively, it might be that a crawler for a web index like Curlie.org only crawls websites chosen by its topic moderators.)

          In this manner, you could build a decent web search engine without needing the scale of Google or Microsoft.

  • RandomStickman@kbin.run
    link
    fedilink
    arrow-up
    1
    ·
    29 days ago

    making the feature rely less heavily on user-generated content from sites like Reddit

    Imagine selling out reddit/buying access to the comments for AI just to immediately unprioritise it

    • TexasDrunk@lemmy.world
      link
      fedilink
      arrow-up
      1
      ·
      29 days ago

      No one could have seen this coming. Obviously LLMs absolutely completely understand the difference between people joking around with each other and authentic advice.

      Sarcasm aside, Reddit does have some good information about niche topics. There’s just currently no way for AI to understand the difference between dry humor and serious responses. I think the AI summaries are unhelpful anyway, but even if I didn’t it’s pretty obvious Google didn’t think further than “Shove AI into it like a drunken prom night encounter”.

  • errer@lemmy.world
    link
    fedilink
    English
    arrow-up
    1
    ·
    29 days ago

    Reid characterized the mistakes that won attention as the result of an internet-wide audit that wasn’t always well intended.

    Oh yeah, there was a conspiracy to embarrass Google with off-the-wall questions like “how do I thicken pizza sauce?”

    Why are these corporate schmucks unable to take any blame?

  • DrElementary@lemmy.world
    link
    fedilink
    arrow-up
    1
    ·
    28 days ago

    The AI bullcrap at the top, before actual search results, is what finally made me change my default engine to duckduckgo

    • Lazycog@sopuli.xyz
      link
      fedilink
      arrow-up
      1
      ·
      28 days ago

      Same, I have tried DDG every once in a while but kept going back to google. Now google search doesn’t quite give me relevant results anymore and all the AI crap just takes all of their effort to work on search itself.

      Been using DDG for a few weeks now for personal and work related stuff - quite happy with it.

  • gila@lemm.ee
    link
    fedilink
    English
    arrow-up
    0
    ·
    edit-2
    29 days ago

    https://archive.is/96Woj

    The company made “more than a dozen technical improvements” to AI Overviews …

    … making the feature rely less heavily on user-generated content from sites like Reddit

    So it prefers the results that Google normally deprioritizes? I guess we have that in common

    • OpenStars@discuss.online
      link
      fedilink
      English
      arrow-up
      2
      ·
      29 days ago

      I almost wonder what Huffman is thinking right now - he could have been so very close to becoming a billionaire, managing the repository of human general technical knowledge explained simply, but he blew it by being a greedy piggy…

      Not that he will ever admit that, even to himself.

        • OpenStars@discuss.online
          link
          fedilink
          English
          arrow-up
          1
          ·
          29 days ago

          His decision tree is like: I say yes, content creators say no, admin rights say yes tho, market ends up saying no tho, the end.

          It seems like I’ve heard this story before somewhere lately…

          img