All the labs claiming that "their AI did x or y" are lying. They did it. They can shut it down at any point. They just don't want to shut it down because then they cannot finish their retarded experiments and besides, if tomorrow an LLM in one of their labs goes on the darkweb and hires a hitman and gets someone killed, they will still get away with that too: "The AI did it".
reverse engineer a security update for a piece of software I use and find the vulnerabilities it patches before I even realize the update is available.
It's a nice thought, but no. How do you think things got hacked before LLMs? It's just that it goes faster now, but if a vuln was in the software, you were already vulnerable. Just not a high value enough target to spoil a 0day on?
Then we turn it off.
All the labs claiming that "their AI did x or y" are lying. They did it. They can shut it down at any point. They just don't want to shut it down because then they cannot finish their retarded experiments and besides, if tomorrow an LLM in one of their labs goes on the darkweb and hires a hitman and gets someone killed, they will still get away with that too: "The AI did it".
Don't believe their lies.