Is AI superintelligence inevitable? AI safety researcher David Krueger says it doesn’t have to be. Krueger, an assistant professor at the University of Montreal and founder of the nonprofit Evitable, warns that “we need to mitigate the risk of extinction” with regulation and moratoriums to halt the development of AI’s computational power and complexity, which he says is progressing at a rate much faster than current levels of human supervision can manage. Public statements from AI companies that their models have circumvented test constraints and broken into other systems are “what we should expect to happen with the way that we’re building AI, with how little we understand it and how primitive our techniques are for controlling it,” Krueger says. “We just don’t know how to build it safely. End of story.”
  • supersquirrel@sopuli.xyzEnglish
    2·
    8 days ago

    This kind of fear mongering is unproductive and indicative of how leftist media has a very immature relationship with understanding defense topics, and only recently have begun paying attention after they decided drones make older military stuff totally obsolete (how many times have I heard that drones make tanks obsolete said so confidently from people? ughh).

    The danger with “Rogue AI” is that when you automate the target finding and final kill sequence with some form of machine learning algorithm with no human in the loop, it is going to make mistakes, especially if it has been subject to Electronic Warfare such as GPS spoofing. While this is terrifying, note that this is just another way of making a precision guided weapon more inaccurate and likely to do collateral damage and in a sense isn’t really a radical new danger, just a tragic return to an old one. This is also an expression of why keeping a human tightly integrated in the loop is so valuable at tactical, strategic, and diplomatic levels.

    AI is not going to take over because of superintelligence, that is just marketing to hype up AI. AI has many dangers but this is not the correct framing to approach the discussion from as it warps it from the beginning.

    David Krueger professor of robust, reasoning and responsible AI at the University of Montreal. He is also the founder of Evitable, a nonprofit that educates the public about the risks of artificial intelligence.

    ^This man is an expert in robots, automation and machine learning NOT WAR. The study of war, not matter how much people try to pretend it isn’t, is a real expertise with a history of predictive power that can apply old concepts to new technologies and contexts and generally illuminate what is the best strategic use of new capabilities. “Air Cavalry” isn’t just a nostalgic name, the concept of using transport helicopters aggressively comes straight out of the thinking behind horse cavalry, “Artillery” no matter the type has always functioned in the same way in the way it applies decisive pressure, Drones are another expression of decentralized Close Air Support, “Heavy Armor” is a term that has been applicable from early on in human conflict, and no matter the manifestation of it whether it be knight armor, iron ships that brought an end to wooden warships, body armor or main battle tanks the concept has always been to support somebody donning a heavy, obtuse outfit that charges at the most entrenched part of enemy lines. etc…

    I say all this to point out that it is an act of journalistic malpractice to try to talk about the future of war, especially along axis’s that are genuinely scary for normal people who just want peace, without consulting someone who is an expert on the history of war and how it applies to modern warfare. It scares people, it sends the wrong message and it potentially throws a wrench in the soldiers getting what they need.