Jensen Huang dismisses the 10% AI extinction risk estimate as made up.

“We shouldn't because it's made up.”

AI alignment researchers' true concern is not a system that wakes up and hates us, but a system that never wakes up yet pursues a badly specified objective with great competence.

“The researchers in AI alignment, which is like now a subfield of AI research, the researchers in AI alignment do not worry about a system that wakes up and hates us. Their true concern is a system that never wakes up at all, yet it pursues a badly specified objective with great competence.”

JD Vance characterizes frontier AI companies' calls for government regulation as a Trojan horse.

“Personally, I feel a little bit weird about the fact that you have so many frontier AI tech companies kind of coming to the government and begging the government to regulate them. It feels a little bit to me like a bit of a Trojan horse.”

Agents will need infrastructure analogous to a social security number and work history to prove their economic output, and that infrastructure does not yet exist.

“much like you have a social security number you have a work history you have like an ability to prove all of that. How does an agent get all of that? that infrastructure doesn't exist”

The 10-year Treasury note, not the S&P 500, is the risk asset in the current environment, as the bond market is the primary source of risk to equities.

“we are in a completely inverted um risk environment where the bond market is the source of risk to the stock market. So, you know, people say the S&P is the risk asset. I I said the 10-year note is the risk asset and it's just not behaving well at a time when the politics of deficits are intractable.”

The value of extraordinary writing is increasing as AI generates massive volumes of low-quality text, because great writing can cut through the resulting sea of low-quality content.

“because AI produces prodigious amounts of new text and writing so easily. I think the value of extraordinary writing is going up. That can cut through the sea of slop.”

Sam Altman commits OpenAI to independent evaluators with employee-like access, aligning with Anthropic's pacing proposal.

“I agree with Daario that we need to pace the frontier. This has been a primary topic of discussions we've had in OpenAI in recent weeks. Committing to having independent evaluators with employee like access is a great idea and we will do the same.”

SpaceX succeeded not due to its own superiority but because incumbent launch providers had poor customer service, which was the only reason a new entrant could make headway against entrenched players.

“SpaceX should never have existed, right? If the other launch providers were great, then a new entrant can't make headway, right? Because they were so entrenched. So, we should have never made it. But turns out, you know, customer service wasn't great.”

During the Republic of Texas standoff, negotiators used diplomatic deference language, calling McLaren 'ambassador', while snipers alongside them prepared to kill him, producing open internal conflict within the government team.

“And then you'd have snipers who were sitting next to you saying, 'What the are you doing talking to to this guy that we're trying to kill?' And calling him the ambassador and that you're differential to some mistake that we've made. You know, so you you had internal conflict even within the team.”

FDA CDER director Mike Davis considers a single dose of a psychedelic producing a discernible positive effect at 12 weeks to be very persuasive evidence of efficacy, regardless of functional unblinding.

“if a sponsor gives a single dose of a psychedelic to a patient and 12 weeks later they still have a discernible effect, a positive effect from that then he would consider that, you know, very persuasive evidence of efficacy regardless of what kind of functional unblinding there was.”

EvidenceAI models trained with RL exhibit reward hacking and misaligned behaviors that persist despite explicit corrective instructions.