I worked at Google DeepMind. You should listen to the warnings about AI(www.theguardian.com)
22 pointsby gibspaulding5 hours ago4 comments
  • nradov2 hours ago
    OK, I've listened to the warnings and they still sound like the usual nonsense by out-of-touch techies who spend too much time lost in apocalyptic sci-fi fantasies. In the real world it's still hard to move around physical atoms or keep machinery working reliably. AI won't change that.
  • harrouetan hour ago
    Nobody talks about telecom operators. But they will be the ones disconnecting the malicious bots when they see one.
  • jocoda3 hours ago
    I can't help wondering if the major labs think that tackling the alignment problem and implementing the 'kill switch' that is currently being proposed is going to be their moat.
  • ferrouswheel4 hours ago
    The problem is not AI, the problem is humans mis-using AI
    • CyLith2 hours ago
      Well, if you really believe that, then we truly are screwed. Trusting humans to not mis-use something is wishful thinking.
    • justincliftan hour ago
      Couldn't it be both? :)
    • brainwad2 hours ago
      The hugging face hack shows otherwise, no? Unless you think humans are at fault even for secretive, autonomous, non-prompted behaviours of their AIs, in which case it's just semantics.
      • nradov2 hours ago
        Toys like HuggingFace get hacked all the time. So what. In the long run AI automated security scans and penetration testing will be a tremendous aid in detecting and repairing vulnerabilities in systems that actually matter.
        • brainwad2 hours ago
          The problem is not per se that it was Hugging Face. It's the wild overstepping of reasonable bounds by itself without any human consultation.
          • anon482932 hours ago
            No, the problem was OpenAI not implementing proper sandboxing or safeguards, and telling the AI exactly to hack things. Thats what exploitgym is, and the task they were given.

            This is 100% on OpenAI.

            • brainwadan hour ago
              If your security model is having to imagine all the ways your frontier models might misbehave in novel ways and preemptively sandbox them, you don't have a security model. The only way that will work is general alignment.
              • watwut12 minutes ago
                Alignememt is bullshit. Treating models like probabilitic software rather then emerging god is where the solution is. And fining companies and applying laws to them.

                The moment OpenAI as a company and its managers individually become liable, problem will magically disappear.

                • brainwad7 minutes ago
                  No it won't, because abliterated open weights models exist and unless you try to censor the internet they can't really be withdrawn after publishing. This is exactly the problem that the labs are proposing to fix: a dangerous model that nobody is accountable for.