When a tech CEO says that Ai is dangerous and that we should limit training for safety, what they are really saying that all AI companies that train models are wasting billions in frontier Ai training that is producing smaller and smaller gains, while Chinese open weight models are still eating their lunch. They should be given a breather without them having to admit that they grossly overpromised regarding AI and AGI on the grounds of "AI safety" because they "could win too much" if they have to keep training and pretending to build datacenters under competitive pressure. If they have to publicly come out and say that their models and tech being just "kinda mid" and not really that useful their stock values will collapse (because they are then just losing billions a day for nothing) and that would mean that US economy itself is built on sand and we will have a massive collapse. AI might not be a real national security issue, but the AI bubble popping is. Saying AI-Danger is just attempt to cool down and maybe diffuse the bubble instead of popping it.

I think they talk like this because they want to turn it into a feverish space race with china so we applaud when the government writes blank checks to the AI companies

Because he doesn't control any of the leading vendors and his company is way behind the curve and losing ground fast. Follow the money and the power (control) to fast track understanding of billionaire behaviour.

As he has lost this race, he'll be backing controls being placed on it, up and until M$ catches up, after which he'll go suddenly quiet.

A.I is an incredibly dangerous existential threat to the species that should absolutely scare any sane person ... but that's not what this is. associating LLMs with AI is pure marketing, stop falling for it.

It's fear-baiting as marketing. Every week there's some story about how dangerous or capable AI is. I wouldn't out it past these companies to stage some event to try and boost hype.

AI in cars, if it gets good enough, will replace a third of workers. It looks very close to good enough

AI that the tech bros expect all knowledge workers out

If it gains the ability to form opinions and direct its action for its own benefit, it's a threat

On the job side if people are idle risk of revolt goes up