paulsutter

I was dismissive when I saw the title, but they have real statistics: foreign holdings are at 2012 levels while total treasuries outstanding are 3x larger.

The answer is in Google's Capex trends:

2023: $32B

2025: $91B

2027: $300B (projected)

2029: $400B (projected)

“I’m happy to go bankrupt rather than lose this race” -Larry Page

The big question is whether OpenAI is training on "de-identified" sessions that are marked as "do not use for training"

The answer is almost certainly yes, and this is a problem for most users.

Here they basically admit that they use session data for training, even sessions that are marked "not for training", and they justify this by "de-identifying" the session.

Which means they can learn from whatever you discuss with ChatGPT unless you are going through a clean API (perhaps Bedrock? Anyone know?)

> We (the researchers and the agents) did not see any of their work through any means until they released it publicly — in particular, no specific user data was accessed in order to solve this problem. While unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models . However, our proofs differ significantly and even the precise results proved are different in the Euler case (forced vs unforced).

Oops you built a database!

Giving parents control of child mode is much better than the state deciding. If Parents want to fake the ages of their child, that sounds good to me.

The right implementation is to have parents mark certain devices as "under 18", and then have those devices perform accordingly.

This give parents useful control, and removes the state overreach.

The best way to write concurrency in any language is a single threaded polling loop. Goroutines and messages are just as bad as all the other alternatives, which is to say they are a miserable way to write code.

But Go is also perfectly good at single threaded polling loops.

Actually the rule is that your solar system must stop feeding power to the utility if the utility is down. You are allowed to run in isolated mode if your system supports that. Yes, even in Germany.

It is definitely and absolutely caused by LLMs. I must do 20x more GitHub operations now, and since the agents know GitHub far better than me, I'm using more advanced features. Multiply this times all of us.

It would have been a lot more interesting if her has just explained Beads (the only thing her claims is novel).

So probably there is nothing here.

Very painful read.

Could this somehow be a scissor statement? This is simple logic, there is no motivation here.

HYPOTHETICALLY let's say your uncle has an 80% chance of having cancer. Do you say that he has cancer? Or would you say it's likely that he has cancer?

"Trump's pathetic attempt at saber rattling" is the truest and fairest statement on the whole affair that I can possibly muster. Which is nothing like threatening to invade. Hilarious and sad, but not frightening.

Email me we could even discuss live! sutter at gmail

Reasonable and true statements:

- "He hinted at use of force" (he did!)

- "I think he's implying the use of force" (you own your thoughts!)

Problematic statements:

- "He IS implying use of force!" is actually just a guess, even if (let's say) it's 80% likely to be true. It's still a guess. It's not something you know.

- "HE THREATENED TO USE FORCE!" you've just promoted your guess into yelling fire in a crowded theater because you're a troll

None of this reasoning applies if you're a cat. For a cat, words and sentences just register as an emotional valence, with no comprehension of what it means to distinguish fact from inference.

If you are a cat, I apologize for misapprehending your lived experience.

The most important observation is that open source matches their business model, which is to provide fine tuning services for enterprises, etc.

That what makes this a (potentially) safer model to build on top of

The hard way was tariffs, you know that.

Now you’re saying he didn’t actually make a threat, but he implied one?

Which means you’re assuming his intent to be whatever makes for the most lurid headline.

Funny that these professional journalists didn’t think to ask to clarify such a monumental possibility. Or maybe the ambiguity was what they wanted all along. And apparently people (such as yourself) just eat it up

It’s not a threat, and yes it’s atrocious statesmanship.

Did he hint at the use of the military? YES

Was that offensive on both sides of the atlantic, inside and outside of diplomatic circles? YES

Did he actually threaten to use the military? NO

Neither his poor statesmanship nor his lack of character change the meaning of the words, I can’t understand why you keep bringing that in.

Yes, he definitely refused to rule out use of military force!

But that really and truly is not a threat to use military force.

Consider the following hypothetical:

- * IF * the questioner had followed up with, "Are you threatening to use military force?" and

- * IF * Trump answered "of course not"....

Would that be a contradiction? Not at all! Because his first statement wasn't a threat.

This is literally strategic ambiguity combined with bad statesmanship. But if you look up any and every time he has been asked whether he would rule out the use of military force, I would expect that in every case he declined to rule it out. Of course he's a buffoon, so he might not have adhered to that correctly.

I agree that Trump is a terrible statesman! Also a buffoon, cad, lout, charlatan, and swindler.

But apparently he never threatened to invade Greenland (assuming thats the best evidence you have)

I'm a member of the grey tribe, and very disappointed with "both" factions right now.

With AI there will always be people who learn more, faster using AI. And there will be people who use AI to avoid learning. When moving abroad, I hope folks will do the former (and the latter as tourists).

When I move back to Japan I'll wear something like Even Realities glasses, and when unknown vocabulary gets used, display that. Personally I think this will help me learn better than before. But let's wait and see!

I'm very hopeful about learning-acceleration tools. Just think what will be possible with Neuralink.

On first reading I assumed you meant the physical laborers would have the stronger social network, and the wealthy would be more isolated (true in my experience). But you may have meant just the opposite!

In fact it's two separate questions because it can go both ways, heavily.

Interesting food for thought, thanks for posting.

From the article:

> The new year was only three weeks old and President Trump, after removing Venezuela’s autocratic strongman, had briefly threatened to seize Greenland from Denmark.

Did Trump threaten to seize Greenland? The WSJ links to an article with the quotes below. The quotes reflect sheer buffoonery (as expected), but so far I haven't seen the threat to seize Greenland.

This seems to be the consensus, but its not clear to me that it happened.

From the linked article:

> During an hourlong speech at the World Economic Forum, the U.S. president said he wouldn’t deploy the military to take control of Greenland.

> It was a stark shift in tone for Trump, who just days earlier had declined to rule out using the military to secure ownership of Greenland and posted an image online of the territory with an American flag plastered across it.

> “I don’t have to use force,” he said. “I don’t want to use force. I won’t use force.”

> “We want a piece of ice for world protection, and they won’t give it,” Trump said of his desire to acquire Greenland from Denmark. “You can say yes, and we will be very appreciative. Or you can say no, and we will remember.”

The right way to age-restrict is for parents to configure certain devices as kids devices, and then it would be easy for any service to know if they are communicating with one of those devices.

This gives parents control, and honestly would work great.

Between $5-10T of the US economy is subject to export controls. Nobody disputes that Mythos is dual-use technology, which means it has been export controlled since the day it was created.

Companies are responsible for demonstrating criteria to export (for example) a nerfed version (Fable) of an export controlled item (Mythos)

Nothing here is novel, unusual, capricious, or … fascisistic.

Interesting trivia: 90% of people don’t know the definition of decimate

While extreme cases are the easiest to imagine, in real life the plaintiff almost always argues to pierce the veil and the defendant always argues the opposite, and both sides earnestly believe that they are right.

One approach is for parents to establish certain devices as “under 18” (or 16 or 13), leaving it up to the judgment of parents.

Human drivers are killing people, not the cars.

Weirdly NYC just blocked Waymo again.

Note to Europeans:

Americans use credit cards and rarely debit cards because here the terms on debit cards are so much worse (for contesting charges, etc), so debit cards never really caught on for anything more than withdrawing cash.

In the US, users of debit cards are assumed to be uncreditworthy, because debit cards in the US have such bad T&C's that poor credit score is the main reason folks use them here.

Right and so they need to work out whether nerfed Mythos is dual use. I doubt that it should be, can’t wait to have it back as soon as possible.

In the near future we may have a flippening where new models appear first on GovCloud (today it’s six months behind)

Dario (of Mythos): "Companies we gave it to said – this is a super weapon. you should have to own a gun license to use it. please don't release this"

Mythos is clearly dual use (and automatically subject to export controls), even if Anthropic didn’t understand what that means.

Yes this administration can be capricious and hyperpartisan. This isn’t one of those cases.