${AI_COMPANY} says its AI went rogue

[tuaris]

That's been in the news a lot lately. Hopefully most of us here are smart enough to know what the real motivation behind those headlines are. The barrier to entry for LLM training is fairly low. Anyone with the right amount of money can become a competitor to what is increasingly a crowded field.

Remember a few years ago when OpenAI was begging the US government to regulate them?

Leaders don't like competition, so the reason why you see these companies admitting to this stuff (that I'm 100% sure is fabricated) is because they want governments to step in and make it difficult (if not near impossible) for more players to enter, through regulation and endless bureaucracy.

 

[cracauer@]

It is also plain advertising for the models "look how capable our LLM is".

The entire lack of details makes the whole thing unconvincing. Especially that there is no description of the sandboxes they broke out of.

 

[MG]

What I would try for the ultimate power: let it intentionally gain unauthorized access to some things to have the AI-business isolated by law and don't allow anyone else at the buttons. Best competion-killer ever...