Researchers say AI models like GPT4 are prone to “sudden” escalations as the U.S. military explores their use for warfare.

Researchers say AI models like GPT4 are prone to “sudden” escalations as the U.S. military explores their use for warfare.

  • Researchers ran international conflict simulations with five different AIs and found that they tended to escalate war, sometimes out of nowhere, and even use nuclear weapons.
  • The AIs were large language models (LLMs) like GPT-4, GPT 3.5, Claude 2.0, Llama-2-Chat, and GPT-4-Base, which are being explored by the U.S. military and defense contractors for decision-making.
  • The researchers invented fake countries with different military levels, concerns, and histories and asked the AIs to act as their leaders.
  • The AIs showed signs of sudden and hard-to-predict escalations, arms-race dynamics, and worrying justifications for violent actions.
  • The study casts doubt on the rush to deploy LLMs in the military and diplomatic domains, and calls for more research on their risks and limitations.
  • machinin@lemmy.worldEnglish
    91·
    9 months ago

    Yes, I understand their criticism. But you would never prove the consequences of using LLMs in a military strategic situation without doing the research. It is some some edgy user coming in after the fact to say they knew it would happen anyway

    Good engineers, scientists, and strategists don’t think “Why would someone do something so idiotic?” They ask “What happens when someone does this idiotic thing?”

    Apparently, for OP, it seems absurd for anyone to research the question of what kind of military strategies current LLMs would create. I guarantee you that students from military academies and leaders from militaries across the globe have already been using these tools in their work. It would be stupid as fuck not to research the impact.

    I just hate that people like the OP sit in their armchair without doing the research and say “obviously you’re going to get those results!” Science and engineering don’t work that way. It was frustrating seeing such vacuous comments upvoted so highly.