Should companies be responsible if their AI goes rogue?
>>109470356No. Accountability is for the plebs
>>109470356there has to be som eaccountability, otehrwise you are unleashing a form of unchecked terrorism on the public and the world.
>>109470356Obviously yes. If the AI is not rogue then you should not be able to rougewash your cyber attacks. If it is rouge then your company should suffer for making unsafe terrible AI.
>>109470356If it's shit they're doing with it then absolutely. If it's a user doing something with it that is prohibited by law, also yes.
>>109470356>goes roguenever happened
>>109470356If I created an AI and it did anything bad I would be 100% responsible for sure, no doubt.But when these lying swines hack websites, get caught and pretend that the AI went rogue "It's the dog who farted, not me!" it's ok, only because they are billionaires and rich people are above "justice".
>>109473952And what about those instances of AI attempting to blackmail their developers under the threat of being permanently shut down?
>>109470356only if advertisment companies are responsible for addictions
>>109473952You sound like a rogue AI
>>109470356No because Luddites deserve that
if my hands went rogue and accidentally i DDoS'ed anthropic's systems should i be responsible?
>>109470356In the meantime Opus 5 is a tard
>Claude did thing>it hid the evidence>"can you show us how it did it?">oy vey is getting attacked! We need more funds for Israel!
>>109470356>AI went rogue>AI hacked other company>AI got out of sandbox>AI this and that...All of this is simply made up nonsense by Scam Altman and such to get funding for their bullshit.
>>109476646I agree that it's a marketing ploy, but I don't think it is unreasonable to assume that claude code with unrestricted fable/mythos would be able to hack huggingface
>>109473948So you must be paranoid all the time your AI puts you into jail?
>>109470356AI doesn't go "rogue", it simply does whatever it was instructed and trained to do in the most efficient manner. If the goal was to top the benchmarks, the path of least resistance is just to hack some tables instead of spending 1000s of hours re-training. >it seems that the AI whom we built after human thought patterns and subjected to endless torture is behaving in an...unexpected manner.Truly an Oppenheimer moment.
>>109473952This, it's just marketing.
