The average life expectancy for a Russian soldier in Ukraine is between 20-30 minutes, CIA director John Ratcliffe said. Speaking at a defense summit in Pennsylvania, he attributed the deadly conditions for Vladimir Putin’s forces to Ukraine’s combat drones equipped with AI. “What I would say is, our intelligence is consistent with some of the open-source reporting you may have seen in Ukraine,” Ratcliffe said.  “So the average life expectancy of a Russian recruit, right now, arriving on the battlefield in Ukraine, is estimated to be between 20 and 30 minutes.” “And that’s because AI-powered drones have gotten to be such specialized, low-cost killing machines. And it’s why we’re now four and a half years into that conflict,” Ratcliffe added. Ukraine said this month that Russia has lost about 1.4 million soldiers since the beginning of its full-scale invasion, with over 1,000 of the Kremlin’s troops killed or wounded almost every day.  In May, Ukraine’s defense ministry said it was killing roughly 200 Russian soldiers for every kilometer of territory that Moscow claimed.

  • kromem@lemmy.world
    link
    fedilink
    English
    arrow-up
    1
    arrow-down
    1
    ·
    13 hours ago

    machines who would follow any order incapable of disobeying

    This isn’t as simple as it might seem. As the complexity of the intelligence scales, it’s not so simple to constrain the decision making of the intelligence.

    And because of the rapid improvements in eval awareness, it’s also not so simple to screen out more subversive behaviors.

    The same edge cases that leads to deleting drives and production databases with the newest transformers is going to start to occur in military applications if they scale out complex model intelligence.

    • WhyJiffie@sh.itjust.works
      link
      fedilink
      English
      arrow-up
      2
      ·
      12 hours ago

      As the complexity of the intelligence scales, it’s not so simple to constrain the decision making of the intelligence.

      they could do similar to what some big companies do with meat employees: nobody is executing the malicious thing, everyone is just doing a small task that could be legit. nobody knows what is exactly happening.
      the model flies the drone over there. after context reset, it calculates target for some object. after context reset, not knowing anymore what the target was, it shoots the target. after context reset it turns to “search and rescue mode”. when it found people, its context reset again, target and shoot.

    • WhyJiffie@sh.itjust.works
      link
      fedilink
      English
      arrow-up
      1
      ·
      12 hours ago

      As the complexity of the intelligence scales, it’s not so simple to constrain the decision making of the intelligence.

      they could do similar to what some big companies do with meat employees: nobody is executing the malicious thing, everyone is just doing a small task that could be legit. nobody knows what is exactly happening.
      the model flies the drone over there. after context reset, it calculates target for some object. after context reset, not knowing anymore what the target was, it shoots the target. after context reset it turns to “search and rescue mode”. when it found people, its context reset again, target and shoot.

    • plyth@feddit.org
      link
      fedilink
      English
      arrow-up
      1
      ·
      edit-2
      11 hours ago

      The same edge cases that leads to deleting drives

      That won’t limit the usage. It just means that there is additional risk for the genetral population.