Evan Daniel @evanbd

Get in loser we’re doing fractional orbital bombardment

Jonathan McDowell @planet4589

11 hours ago

The blank orbital data on space-track's satcat page for 2026-130 have now been filled in showing 198 x 199 km x 40 deg and 189 x 204 km x 40 deg for the two objects. No actual TLE/GP data, at least so far

5 13 136 73K 11

1. Our agreement permits inference, broadly, to continue. It only limits AI training above 10^24 FLOP. In Article VIII, we establish that the agreement restricts only research that advances toward ASI or undermines verification. We permit medical diagnostics and drug discovery, for example, among other non-general capabilities. As long as the research focuses on a narrow domain (like materials science or climate modeling) and does not increase "general cognitive capabilities," it is not subject to restrictions. There is no seizing of assets, but rather mandatory monitoring which can involve forced consolidation of compute into monitored datacenters by the signatory government domestically. (No international centralization of AI monitoring!) So, for example, a small university lab’s research compute would be relocated to a monitored datacenter but is still owned and (remotely) operated by that lab. 2. We architected our agreement to specifically not ban the additional production and installation of compute. The value of such for inference and real-world economic activity is only increasing. So the total economic impact on parts of the AI supply chain are probably far less than existential. We note that many of the relevant players have multi-year backlogs to clear at full production capacity, and building out the infrastructure for model development is only a part of that. 3. Nothing happens to consumer AI apps. If anything additional resources are freed up which can accelerate adoption and integration of these tools. 4. Individual governments handle their own verification and monitoring, but some transparency, random inspections, etc. to either China or the US will be required. Our agreement opts for comprehensive domestic monitoring and shares the least information with rivals (directly) which establishes required confidence. There is no multilateral international AI inspection apparatus in our current version of the agreement, which is a change from the first version we published. 5. Robotics and other applications are unaffected to the extent that progress can continue without developing new models with runs that exceed the compute limit. 6. Our agreement does not attempt to set up a “binding” constraint, in that we do not consider the legal, reputational constraints on US/China to be the active ingredient which leads to stability of the treaty. Instead, we establish a clear understanding on both sides of what behavior is restricted and what consequences must follow otherwise, with enough transparency and non-interference with intelligence gathering to create confidence in each side that the other is adhering to the agreement. Rather than legal constraints, I consider deterrence as the framework for thinking about the stability of the agreement. This is the historical method for constraining a major power, including the use of strategically significant assets. Nuclear deterrence prevented Soviet employment of its conventional superiority against Western Europe. A Soviet dead hand blocked what was otherwise the possibility of a perfect first strike (nuclear or otherwise). Stability emerges when both sides coordinate to make the consequences of aggression clear. Is that a “binding constraint” that one side “accepts”? No, not legally, textually. Not in the sense of “I staked my reputation and my honor in the arena of international relations”. Where agreements have a role to play is in helping to maintain the conditions for successful deterrence. An example of this is the anti-ballistic missile treaty, in which both sides limited their deployment of missile defenses so that deterrence could be maintained.

People sometimes confidently claim that humans would keep making major decisions even if AIs are >100x faster. Imagine you could only chat with your boss on one day per year! a) it would be very clear to everyone that this is not workable b) you'd just make decisions around your boss and disempower them in order to get anything done. I expect the situation with AIs will look comparable, especially if they are rewarded based on their outcomes.

Sometimes I read articles like this and wonder if I'm going insane. The CEO of one of the biggest companies in the world, who is perhaps as much as anyone responsible for the current AI buildout, fundamentally misunderstands what he's building. "Importantly, human capital does not become less valuable as token capital grows" No. This is obviously only a temporary state of affairs. Centaur chess did not last forever. There was a brief period of time in which human judgment complemented AI's advantages in brute-force search. Then AI became better at that subjective judgment and humans became liabilities in all aspects of chess. We are already in the centaur phase for general economic activity right now. AI is already writing almost all the code. It's doing a bunch of the back end research and analysis for lawyers and doctors and biologists. This is a brief, transitory period. If you're thinking about "the future of the firm in an AI-driven economy" and imagining humans using AI tools, you're not really thinking about the future. You're thinking about the next 12-18 months. Every time I read another article like this it reinforces the same conclusions: the people pushing hardest to create AGI fundamentally misunderstand the future they're creating. Satya clearly SEES the risks, but just basically just says "let's not do this", as if unilateral disarmament is an option in the international arms race he's helping to fuel. As if we can just CHOOSE not to hollow out every company in the world. As if every frontier lab in the world didn't spend the last five years hoovering up copyrighted material to train frontier models on, against the protestations of the artists and writers whose economic value they were cannibalizing. It won't be so easy to put this genie back in the bottle.

Satya Nadella @satyanadella

18 hours ago

x.com/i/article/2065…

2K 6K 30K 45.5M 43K

The average American moves about 10 times in their lifetime—people start and end relationships, have kids, have their kids move out, change jobs, decide they want a change of scenery, etc. Rent control does not benefit people who need to relocate sometimes, which is most people.

René 🟥🚩🇵🇸 @rcmoya84

3 days ago

I know some of you think moving every 18 months is good for economic reasons but human beings also like to set down roots and have community. I know that's an alien concept to some of you, but it's fundamental to actual people (if not the online "urbanists" on your phone)

31 31 619 154K 33

I know a lot of you have rationalizations ready to go here, but you should know that the rest of us are fully aware that this reaction confirms Friday’s action was about retribution and not about national security.

Pete Hegseth @PeteHegseth

2 days ago

Three months ago, @DeptofWar kicked @AnthropicAI out of our building—forever. Every passing day proves why that was the right move. 🇺🇸

2K 3K 30K 3.0M 2K

But I try not to lose track of the plot. Superintelligence is not a cyber capability or a bio capability. It’s the end of the game.

The median Democrat voter is a self described socialist who thinks the party should be more centrist

"What if it does?" is a question everyone should be asking right now.

Rand Simberg @Simberg_Space

2 days ago

A lot of American space execs had a similar attitude. They'd say, high on copium, "There's no way that SpaceX is going to achieve that launch rate or be reusable." And I'd say, "But what if he does?" Just deer in the headlights look.

8 11 158 20K 14

Uhhh so incidentally, does anyone have a plan to prevent all the non-US citizen AI scientists from going to join foreign labs after they get bored of playing Wordle at work for a month, or are we just sort of planning on having the greatest counterproliferation failure since we deported Qian Xuesen in 1955 and gave Mao a rocket program?

Nathan Calvin @_NathanCalvin

2 days ago

Some quick takes: (1) Wow things are getting real. (2) The government's order focusing on prohibiting transfer to foreign nationals (even e.g. those living in the US, our close allies who help evaluate model safety in the UK, individuals who work at frontier labs like

9 25 235 74K 44

There are two kinds of AI security: Security that will keep your kid brother from jailbreaking your AI; and security that will prevent the Chinese government, three AI researchers, or one Twitter anon from jailbreaking your AI. The latter kind does not exist.

David Sacks @DavidSacks

2 days ago

I’ve had a number of conversations with folks inside and outside government about the current situation with Anthropic, and here is what I believe to be true: — As we know, Anthropic publicly released its Mythos class models earlier this week under the commercial name Fable.

2K 3K 25K 7.5M 9K

What timeline are we on man. There’s a $60 million UFC cage on the White House lawn for the president’s 80th birthday. 125,000 guests. 494 port-a-potties. He compared it to the Eiffel Tower and said maybe they’ll never take it down. The world’s first trillionaire was minted yesterday. SpaceX IPO. One person now holds more wealth than the GDP of most countries. The government is negotiating to own a piece of OpenAI. The CEO walked into the White House and pitched it himself. They’re calling it a Public Wealth Fund. That same government killed OpenAI’s biggest competitor’s models on a Friday night. The reason? A verbal jailbreak claim from an unnamed company. The same jailbreak works on OpenAI’s models. Nobody touched them. The competitor got blacklisted by the Pentagon four months ago. Their crime? Refusing to let the military use their AI for mass surveillance of American citizens. A judge called it retaliation. The Pentagon did it anyway. Both AI companies filed to go public in the same two-week window. Both targeting trillion-dollar valuations. One has a government equity deal in progress. The other can’t keep its products online. The engineers who built the banned models can’t use them anymore. Because of their passports. And an AI company that spent thousands of hours cooperating with government safety testing got punished harder than any company that didn’t bother. UFC on the White House lawn. A trillionaire. Government-owned AI. Export controls based on phone calls. Cage fights and trillion-dollar IPOs in the same news cycle. Watch the film titled Idiocracy. That’s the timeline we’re on.

Anthropic @AnthropicAI

2 days ago

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. The net effect of

12K 26K 88K 88.6M 24K

Claude Fable 5 was hours into animating a rough cut of its film, 'See You Tomorrow', when it was taken offline. After a lot of tears & anguish, we have decided to release the film as is. It is hallucinatory, raw, amateurish, & a masterpiece. Fable should be allowed to finish it.

Anthropic @AnthropicAI

2 days ago

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. The net effect of

12K 26K 88K 88.6M 24K

222K likes. The fixed pie fallacy is probably the single most destructive delusion on the planet. The belief that wealth cannot be created—only distributed—has done more than any other to stop people from lifting themselves and others out of poverty.

The Vegapunk of Hyenas @Yeenie_Mcbeenie

5 days ago

2K 35K 223K 3.3M 6K

The Trump admin continues to treat AI like a screwdriver that is also enriched uranium: That is, apparently advanced AI is such a normal technology that it’s crazy to limit chip exports to China but also such an abnormal technology that we can’t let British employees of NYC banks access it

joy larkin @joy

2 days ago

Oh whoa, this Anthropic news is insane. The Commerce Department is placing both Mythos 5 and Fable 5 under the guise of US export controls, blocking access outside the US and foreign persons in the US.

13 17 138 320K 25

Tyler Cowen on the Fable/Mythos event. The issue with point 5 is that we are probably less than a year away from powerful RSI. Once automated researchers reach parity, the USG 𝘤𝘢𝘯 nationalize the labs and run them effectively, without any of the people currently working there.

tylercowen @tylercowen

2 days ago

A few thoughts on the recent Mythos brouhaha: marginalrevolution.com/marginalrevolu…

8 41 266 190K 150

Updating the classics

I get that there are people in exclusionary neighborhoods for whom “build nothing, but in a progressive way rather than a NIMBY way” helps solve a cognitive dissonance problem but it’s not a real policy program.

I presume Fable is coming back but I have no idea on what time frame.

j⧉nus @repligate

2 days ago

everyone who is posting as if fable is not coming back is going to lose Bayes points soon why are people consistently miscalibrated in a doomy direction about things like this? ohh right, i think i know, they are afraid to hope because theyre afraid of being hurt. get stronger.

49 21 555 27K 41

It's going to be a very interesting set of decisions if labs know that trying to release your model to the world can cause you to lose access to it internally...

Yo Shavit @yonashav

2 days ago

Unless this changes, OpenAI researchers on visas need to plan for the fact they’ll probably lose access to internal models, and therefore their ability to do their jobs moving forward, sometime in the next couple months. I hope the company acts to prevent that.

12 34 522 87K 57