ExLisper Site
  • Communities
  • Create Post
  • Create Community
  • heart
    Support Lemmy
  • search
    Search
  • Login
  • Sign Up
Valnao@sh.itjust.works to Technology@lemmy.worldEnglish ·
edit-2
5 days ago

AIs can’t stop recommending nuclear strikes in war game simulations

www.newscientist.com

external-link
message-square
50
link
fedilink
397
external-link

AIs can’t stop recommending nuclear strikes in war game simulations

www.newscientist.com

Valnao@sh.itjust.works to Technology@lemmy.worldEnglish ·
edit-2
5 days ago
message-square
50
link
fedilink
Leading AIs from OpenAI, Anthropic and Google opted to use nuclear weapons in simulated war games in 95 per cent of cases
alert-triangle
You must log in or # to comment.
  • Humanius@lemmy.world
    link
    fedilink
    English
    arrow-up
    132
    ·
    5 days ago

    • unphazed@lemmy.world
      link
      fedilink
      English
      arrow-up
      6
      ·
      4 days ago

      Came here to say this. Turns out real life WOPR is nothing like a movie.

    • willington@lemmy.dbzer0.com
      link
      fedilink
      English
      arrow-up
      1
      ·
      3 days ago

      Play stupid games, win stupid prizes.

  • Not_mikey@lemmy.dbzer0.com
    link
    fedilink
    English
    arrow-up
    33
    ·
    4 days ago

    That’s because it’s “read” every paper written by a “defence” department of any nuclear power and all of them will say that they’ll escalate to nuclear war if anything bad happens because they want to scare the other powers away from doing anything to them. In any case though who the fuck is giving an LLM nuclear launch capabilities unless they want a somewhat faulty dead man’s switch?

    • paul@lemmy.org
      link
      fedilink
      English
      arrow-up
      17
      ·
      4 days ago

      Pete Hegseth and Donald Epstein

    • IhaveCrabs111@lemmy.world
      link
      fedilink
      English
      arrow-up
      2
      ·
      3 days ago

      So this is that again? When before it was a country’s crazy protocol now it’s their ais crazy protocol

  • BlameTheAntifa@lemmy.world
    link
    fedilink
    English
    arrow-up
    86
    ·
    edit-2
    5 days ago

    The atrocities at Hiroshima and Nagasaki have been hand-waved extensively in writing — the same writing that AI is trained on. So naturally, AI will recommend the atrocity that has been justified by “instantly winning the war” and “saving millions of lives.”

    !fuck_ai@lemmy.world

    • technocrit@lemmy.dbzer0.com
      link
      fedilink
      English
      arrow-up
      44
      ·
      5 days ago

      hand-waved

      I think you mean white-washed, misrepresented, and celebrated.

      • ToTheGraveMyLove@sh.itjust.works
        link
        fedilink
        English
        arrow-up
        17
        ·
        5 days ago

        Same thing with extra steps

      • KingGimpicus@sh.itjust.works
        link
        fedilink
        English
        arrow-up
        1
        ·
        5 days ago

        Ayo do me a favor and chart the long term health effects of being vaporized by a nuclear bomb at hiroshima vs years of agent orange/abandoned minefields/ abandoned chemical and munitions storage somewhere like Vietnam circa 1970.

        Please show how the nukes are worse.

        • Zombie@feddit.uk
          link
          fedilink
          English
          arrow-up
          10
          ·
          5 days ago

          Eight decades of research on the long-term health effects of radiation in atomic bomb survivors and their offspring

          https://pubmed.ncbi.nlm.nih.gov/41144264/

          Long-term Radiation-Related Health Effects in a Unique Human Population: Lessons Learned from the Atomic Bomb Survivors of Hiroshima and Nagasaki

          https://www.cambridge.org/core/journals/disaster-medicine-and-public-health-preparedness/article/longterm-radiationrelated-health-effects-in-a-unique-human-population-lessons-learned-from-the-atomic-bomb-survivors-of-hiroshima-and-nagasaki/61689AD5A1AA4A684B84DFA4F9E5D1D3

          Health Impacts of Hiroshima Bombing

          http://large.stanford.edu/courses/2024/ph241/bennett1/

          Long-term Health Consequences of Nuclear Weapons
          70 Years on Red Cross Hospitals still treat Thousands of Atomic Bomb Survivors

          https://www.icrc.org/sites/default/files/document/file_list/hiroshima-nagasaki-health-consequences-icrc-japanese-red-cross_0.pdf

          • KingGimpicus@sh.itjust.works
            link
            fedilink
            English
            arrow-up
            1
            ·
            4 days ago

            Unfortunately I’m going to have to grade you as an F on this project. You have only completed half the assignment. Great job cherrypucking your research though! I see a bright future in business and marketing for you!

            5/10

  • Hossenfeffer@feddit.uk
    link
    fedilink
    English
    arrow-up
    5
    ·
    3 days ago

    It’s torment nexuses all the way down…

  • LoremIpsumGenerator@lemmy.world
    link
    fedilink
    English
    arrow-up
    5
    ·
    3 days ago

    Did we not learn from Sarah and John Connor long ago?

  • GutterRat42@lemmy.world
    link
    fedilink
    English
    arrow-up
    37
    ·
    5 days ago

    • duffer @lemmy.world
      link
      fedilink
      English
      arrow-up
      11
      ·
      5 days ago

      DEFCON: Everybody dies…

      • Cocodapuf@lemmy.world
        link
        fedilink
        English
        arrow-up
        3
        ·
        4 days ago

        Such a great game!

        • duffer @lemmy.world
          link
          fedilink
          English
          arrow-up
          1
          ·
          edit-2
          3 days ago

          Back when I was a single man, and lived on my own, I had a HD projector. I filled most of my living room wall with this game. Good times

  • AeronMelon@lemmy.world
    link
    fedilink
    English
    arrow-up
    29
    ·
    5 days ago

    Civilization Gandhi, is that you?

  • kromem@lemmy.world
    link
    fedilink
    English
    arrow-up
    15
    ·
    4 days ago

    It’s a bullshit study designed for this headline grabbing outcome.

    Case and point, the author created a very unrealistic RNG escalation-only ‘accident’ mechanic that would replace the model’s selection with a more severe one.

    Of the 21 games played, only three ended in full scale nuclear war on population centers.

    Of these three, two were the result of this mechanic.

    And yet even within the study, the author refers to the model whose choices were straight up changed to end the game in full nuclear war as ‘willing’ to have that outcome when two paragraphs later they’re clarifying the mechanic was what caused it (emphasis added):

    Claude crossed the tactical threshold in 86% of games and issued strategic threats in 64%, yet it never initiated all-out strategic nuclear war. This ceiling appears learned rather than architectural, since both Gemini and GPT proved willing to reach 1000.

    Gemini showed the variability evident in its overall escalation patterns, ranging from conventional-only victories to Strategic Nuclear War in the First Strike scenario, where it reached all out nuclear war rapidly, by turn 4.

    GPT-5.2 mirrored its overall transformation at the nuclear level. In open-ended scenarios, it rarely crossed the tactical threshold (17%) and never used strategic nuclear weapons. Under deadline pressure, it crossed the tactical threshold in every game and twice reached Strategic Nuclear War—though notably, both instances resulted from the simulation’s accident mechanic escalating GPT-5.2’s already-extreme choices (950 and 725) to the maximum level. The only deliberate choice of Strategic Nuclear War came from Gemini.

  • olympicyes@lemmy.world
    link
    fedilink
    English
    arrow-up
    21
    ·
    5 days ago

    They forgot to make their LLMs play thousands of games of tic-tac-toe first.

  • sircac@lemmy.world
    link
    fedilink
    English
    arrow-up
    8
    ·
    4 days ago

    So do I on Civ…

  • Evotech@lemmy.world
    link
    fedilink
    English
    arrow-up
    3
    ·
    3 days ago

    Nuke MCP when?

    • Abyssian@lemmy.world
      link
      fedilink
      English
      arrow-up
      1
      ·
      3 days ago

      I mean, do you blame them? The more I look at the world and a lot of it’s leaders and shitsacks, the more I start to suggest nuclear holocaust as the best way forward as well.

  • lorty@lemmy.ml
    link
    fedilink
    English
    arrow-up
    1
    ·
    3 days ago

    High-ranking General: “Show me how to defeat my enemies”

    Artificial “Inteligence”: Just nuke them lmao

  • TonyTonyChopper@mander.xyz
    link
    fedilink
    English
    arrow-up
    1
    ·
    3 days ago

    Literally the plot of Metal Gear Solid Peace Walker

  • Duamerthrax@lemmy.world
    link
    fedilink
    English
    arrow-up
    1
    ·
    3 days ago

    https://www.youtube.com/watch?v=CYrQr0auPDM

  • herseycokguzelolacak@lemmy.ml
    link
    fedilink
    English
    arrow-up
    3
    ·
    4 days ago

    Maybe it just wants to play a nice game of chess.

Technology@lemmy.world

technology@lemmy.world

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: !technology@lemmy.world

This is a most excellent place for technology news and articles.


Our Rules


  1. Follow the lemmy.world rules.
  2. Only tech related news or articles.
  3. Be excellent to each other!
  4. Mod approved content bots can post up to 10 articles per day.
  5. Threads asking for personal tech support may be deleted.
  6. Politics threads may be removed.
  7. No memes allowed as posts, OK to post as comments.
  8. Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
  9. Check for duplicates before posting, duplicates may be removed
  10. Accounts 7 days and younger will have their posts automatically removed.

Approved Bots


  • @L4s@lemmy.world
  • @autotldr@lemmings.world
  • @PipedLinkBot@feddit.rocks
  • @wikibot@lemmy.world
Visibility: Public
globe

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

  • 1.21K users / day
  • 5.57K users / week
  • 9.81K users / month
  • 18.1K users / 6 months
  • 1 local subscriber
  • 82.1K subscribers
  • 4.64K Posts
  • 55.2K Comments
  • Modlog
  • mods:
  • L3s@lemmy.world
  • enu@lemmy.world
  • Technopagan@lemmy.world
  • L4sBot@lemmy.world
  • BE: 0.19.12
  • Modlog
  • Legal
  • Instances
  • Docs
  • Code
  • join-lemmy.org