No. Has the underlying technology changed in any way?
Uh, our understanding of it has. Doing automated cyberattacks using novel zero-days, or resolving open math questions, was perhaps one of the many somewhat plausible futures in 2022, but absolutely not an inevitable or even particularly likely one. If you disagree, I’d like to see your analysis and predictions from four years ago, otherwise it’s nothing but hindsight bias.
Our understanding perhaps. But do you think researchers working on it had a similar understanding?
Ah yes, because white hats aren’t allowed to use AI to find zero days, and solving open math questions is such a horrible thing, why even bother advancing mathematics at all if someone can’t spend ten years and hundreds of thousands of dollars to only get partway there and feel good about themselves? This new understanding definitely makes me want to regulate AI so only the Right People™ can use it, that definitely won’t empower rival nations who lack such regulation and give them an advantage.