Lemmy.one
  • Communities
  • Create Post
  • heart
    Support Lemmy
  • search
    Search
  • Login
  • Sign Up
huf [he/him]@hexbear.net to Chapotraphouse@hexbear.netEnglish · 12 days ago

Some LLMs now outperform Biden

clocks.brianmoore.com

external-link
message-square
30
fedilink
  • cross-posted to:
  • programmer_humor@programming.dev
  • hackernews@lemmy.bestiver.se
91
external-link

Some LLMs now outperform Biden

clocks.brianmoore.com

huf [he/him]@hexbear.net to Chapotraphouse@hexbear.netEnglish · 12 days ago
message-square
30
fedilink
  • cross-posted to:
  • programmer_humor@programming.dev
  • hackernews@lemmy.bestiver.se
AI World Clocks
clocks.brianmoore.com
external-link
The current time as rendered by 9 different AI models. By Brian Moore.
  • Johnny_Arson [they/them]@hexbear.net
    link
    fedilink
    English
    arrow-up
    16
    ·
    12 days ago

    Meanwhile on deepseek:

    Perfectly displaying my local time in real time indistinguishable from an actual clock. China stays winning.

    • juniper [none/use name]@hexbear.net
      link
      fedilink
      English
      arrow-up
      15
      ·
      12 days ago

      The Qwen model is also Chinese data-laughing

      • Johnny_Arson [they/them]@hexbear.net
        link
        fedilink
        English
        arrow-up
        12
        ·
        12 days ago

        Lol. Wonder what went wrong there.

        • BanMeFromPosting [none/use name]@hexbear.net
          link
          fedilink
          English
          arrow-up
          9
          ·
          12 days ago

          Easy. It’s the same people behind both LLMs. They just told a developer to code a program that makes an app that designs an algorithm that makes a neural network that puts all the bad stuff into one LLM and all the good stuff into another and that’s how you get qwen

        • juniper [none/use name]@hexbear.net
          link
          fedilink
          English
          arrow-up
          6
          ·
          12 days ago

          Not enough good training data mao-red-book

        • piccolo [any]@hexbear.net
          link
          fedilink
          English
          arrow-up
          4
          ·
          12 days ago

          IIRC Qwen 2.5 is old and a much smaller model than DeepSeek 3.1, but also a lot of the models are outdated. I know that Qwen makes models that are more comparable to DeepSeek.

        • gay_king_prince_charles [she/her, he/him]@hexbear.net
          link
          fedilink
          English
          arrow-up
          2
          ·
          12 days ago

          A temperature greater than zero

Chapotraphouse@hexbear.net

chapotraphouse@hexbear.net

Subscribe from Remote Instance

Create a post
You are not logged in. However you can subscribe from another Fediverse account, for example Lemmy or Mastodon. To do this, paste the following into the search field of your instance: !chapotraphouse@hexbear.net

Banned? DM Wmill to appeal.

No anti-nautilism posts. See: Eco-fascism Primer

Slop posts go in c/slop. Don’t post low-hanging fruit here.

Visibility: Public
globe

This community can be federated to other instances and be posted/commented in by their users.

  • 757 users / day
  • 1.36K users / week
  • 2.11K users / month
  • 3.87K users / 6 months
  • 17 local subscribers
  • 14.3K subscribers
  • 22.5K Posts
  • 343K Comments
  • Modlog
  • mods:
  • LENINSGHOSTFACEKILLA [he/him]@hexbear.net
  • corgiwithalaptop [any, love/loves]@hexbear.net
  • a_little_red_rat [he/him, comrade/them]@hexbear.net
  • khizuo [ze/zir]@hexbear.net
  • thelastaxolotl [he/him]@hexbear.net
  • CoolerOpposide [she/her]@hexbear.net
  • BE: 0.19.7
  • Modlog
  • Legal
  • Instances
  • Docs
  • Code
  • join-lemmy.org