AI safety

The president says he's, quote, seriously considering that diesel export ban.

Our roadmap begins with NVIDIA, of course, releasing a new AI safety software platform, setting safeguards to prevent agents from breaking out of containment.

Speaking of AI, Anthropix's Dario Amore joining in the president for a White House dinner.

the first one-on-one meeting between those two leaders.

And then there is SpaceX launching the massive Starship rocket moments ago, reaching orbit with the spacecraft for the first time.

Let's begin, though, with NVIDIA up in the pre-market, announcing plans to boost its stock repurchase plan by a record-setting $150 billion.

Meantime, the company released some new AI safety software today, which it says could have stopped hugging face.

This is what Jensen Wong had to say about it last hour on Squawk.

yeah yeah who knows who knows where does that come

from there's been a lot of news over the weekend about the pace of model development and ai safety i suppose you could call it that and then frontier labs so that's the people like anthropic and open ai and some of the others the people making these new models you know releasing them every week and things like that they've kind of come out and they're kind of starting to news out to say we've creating codes of conduct and we think we need to kind of slow down some of the development of this now who knows what's going in the background but what seems to have kind of also started to light that fire we had um new york city you know coming out uh in the news episode two weeks ago where they said they were going to put the moratorium on ai development for kids but at the same time there's been some technology news and noise that have come through this.

So, for example, this week we've had an ex-anthropic insider telling CNN that AI could kill humans by 2030.

We've had OpenAI insider talk about the current OpenAI capabilities and that we should be calming down on the kind of development of OpenAI models.

I'm Emile Torrance.

Today, we're just here by ourselves because the conversation about AI safety seems to have really exploded in the past few weeks here, starting with the hogging face incident and then that guy, Jacob Coxon.

fake whistleblower, fake ass whistleblower in my opinion, whistleblower who didn't blow the whistle on anything, quit Anthropic.

And then also Dario Amadei came out and did his whole routine about like, oh, this is so dangerous.

I was noticing that a couple news programs featured a, like, a talk from that guy.

What's his name? Daniel Cocatalo.

And he's just on there talking about AI safety.

And I'm like, I feel like people don't understand AI.

And I think we might be approaching recursive self-improvement with the more advanced AI models that exist in these frontier labs, where the AI starts to be responsible for an increasingly large fraction of the intellectual work going into designing the next generation of these AI systems.

And I do think that many people in these labs are seriously concerned about the rapid strides being made and whether AI safety will be able to keep pace with AI capability, especially if developments start to accelerate

even further.

So you basically take it more at face value then.