seewhydee

Bypassing the CUDA moat is, in fact, one of the major tasks Deepseek set for itself. Their efforts in this area are likely one of the main reasons for their slow release cadence, culminating in their v4 inference setup that runs on Huawei Ascend chips.

That wouldn't explain why Deepseek is fast relative to other Chinese providers, especially considering that they're reportedly ahead of the curve among Chinese companies in moving off Nvidia. I think their quant fund background has more to do with it. Their models are clearly designed with performant inference clearly in mind.

Had they wanted a good ML relevant physics Nobel, the committee had decades to award a prize to Marshall and Arianna Rosenbluth for the Markov Chain Monte Carlo method. Would have been self-evidently important and relevant to both physics and ML. Too late now -- Arianna died in 2020.

Reducing the levels of allergens in food very likely has the side effect of promoting food allergies in children, due to lack of exposure during development. That's a strong negative consequence.

Not sure how this is "dishing it out and not taking it". They intend to "take it"; they will force Bytedance not to sell, and thereby lose its US business with no compensation.

I am old enough to remember US internet companies spamming users to support net neutrality, even those who aren't American and can't do anything about it. And net neutrality was heck of a lot less of an existential issue for them.

Beijing has a veto on the sale, and are likely to exercise it. From their point of view, allowing the US to force a Chinese company to sell itself off as soon as it achieves success would set a very bad precedent, and encourage a torrent of other blackmail efforts.

They'll be willing to let Bytedance take the L on this. Bytedance can console themselves that TikTok can still operate elsewhere on the world, just making less money.

The trouble with doing that for Freakonomics is that the work on abortions reducing crime, which has been proven wrong, is the first chapter and the centerpiece of the book. It's the thing that they use to exemplify the "freakonomics" approach in the rest of the book.

> Singapore operates at 15% but doesn't really try to field a military

Singapore is one of the most militarized countries on the planet. For example, they currently operate more jet fighters than Australia (100 vs 94).

> Xi’s security-first state employs Orwellian surveillance systems that vastly complicate spy operations inside the country.

Grapes are probably real sour, anyway.

I mean, Plato argued for governance by philosopher-kings, not too different from the Confucian conception of governance.

What's more interesting to me is the skeptical attitude toward omens and the notion of heavenly will. That sounds really remarkably modern, and not at all what I'd expect from someone writing during the third century BCE.

The desktop client is basically a web app, little different from running it in the browser. As for the app, I have been pretty unimpressed by it; performance is sluggish, and emails are displayed poorly (lots of unreadably small fonts with no options for text resizing). Worst of all, development seems to be either inactive, or so slow as to be undetectable.

I paid for Tutanota and started switching to it from Gmail, but the accumulation of inconveniences is starting to make me consider switching back, in spite of all the Google privacy issues.

From Fig. 2, it looks like the "hazard ratio" is minimized at a total cholesterol level of around 230 mg/dL, with both higher and lower total cholesterol levels associated with higher mortality. However, medical advice (see e.g. Cleveland Clinic page below) routinely asserts 200 mg/dL as the maximum for the normal range, with 230 mg/dL being "borderline high" and approaching the "high" band.

https://my.clevelandclinic.org/health/articles/11920-cholest...

In the discussion section of the paper, the authors write:

"The current cholesterol guidelines are heavily based on heart disease risk and recommend a TC range of <200 mg/dL as desirable. TC range <200 mg/dL, however, may not be necessarily a sign of good health when other diseases are considered. The diseases associated with lower TC levels and potential mechanisms have not been conclusively identified."

Time to eat more cheeseburgers?

It stands to reason that someone who can create his own file system can't possibly be guilty of murder.

That German lab should sell LK-99 crystals. I wouldn't mind buying a souvenir for this whole episode!

No, they were not careful about their claims. Putting aside the drama around the arxiv postings, they had previously explicitly made the "discovered a room temperature ambient pressure superconductor" claim in a Korean language journal, in a 2020 submission to Nature that was rejected, and on multiple patent applications.

There is no evidence of them doing interesting work, either. If reporting is to be believed, their company mostly does unrelated consulting odd-jobs for the chaebols, and this was a passion project for Lee and Kim. But you know what? I'm glad there are oddballs like this on the fringe of science. They're mostly harmless, occasionally entertaining, and maybe once in an epoch they might come up with something real.

I'd add that the 3-4 preprints that put the final nails in the coffin for LK-99 --- the ones that showed how ferromagnetic samples produce half-levitation, pointed out the Cu2S structural phase transition, explaining why the flat bands are not conducive to superconductivity, and the creation and characterization of a pure single crystal --- were all good showcases of scientific ingenuity.

It was particularly impressive to see some groups putting together high-quality, publication-ready preprints from scratch in the span of a month. Especially crazy when you think about the shoddy original papers that the Korean group spent years working on...

Yes, I noticed that this article disproportionately quoted American scientists who played relatively minor roles in the replication efforts. I guess the reporter just found it more convenient to reach out to them for comments.

The thing is that there are now multiple independent lines of investigation pointing to LK-99 not being a superconductor, and explaining away the original "smoking guns" offered by the authors.

It's like we have a murder suspect, the murder weapon, and fingerprints lifted from the scene. At this point it could still be space aliens, but nobody in their right mind would treat that possibility seriously.

The video really rubbed me the wrong way. I guess it's a persona he's putting on for the YouTube channel, but that "tough minded skeptic" bit is way over the top. He spent all his time criticizing various problems with the LK-99 paper that were indeed problematic (and widely commented on elsewhere), but didn't necessarily falsify the claim (e.g. they might have come from having mixed phase samples).

And he did not talk at all about the scientifically most interesting part of the affair, which was the clever investigation from multiple angles that finally unearthed the explanation. It's as though he just ran his mouth without reading the literature... which is not a very scientific thing to do, is it?

Yeah. I would argue that he displayed some bad science of his own. He kept harping on about how the LK-99 paper didn't show resistivity going down to exactly zero, but had a residual resistivity. But that is exactly what you would have seen with a mixed phase sample with some superconductor and some normal material. It took proper scientific detective work to distinguish the possibilities, not this kind of silly snark.

Yeah, it needs to be doped with a room temperature superconductor, lol.

> According to some of the new data, the resistivity of LK-99 increases as you lower the temperature, like some sort of anti-superconductor.

The writer based this on a sarcastic tweet, taking it at face value. It seems to imply that resistance increasing with temperature is somehow startling; actually, it's how semiconductors behave (as the original tweeter knows very well).

Already woefully out of date by the time the video was put out.

I would argue that CMTC is acting here more like a twitter drama queen than a proper scientist.

That's a pretty condescending take on what's evidently a paintakingly put-together theoretical+experimental study, from one of the best condensed matter groups in the world.

Believe it or not, researchers more than any other people love the possibility of previously-undiscovered exotic phenomena; these guys are not coming into this with the agenda of enforcing orthodoxy and stifling a potential discovery. But materials still ought to be consistent with what we know about the laws of physics, and it's totally valid to check this via theory and experiment.

Contrary to the linked tweet, there's some reason to think its diagmagnetism is pretty interesting. It might be the second most diamagnetic non-superconductive material known. Why this may happen may keep condensed matter physicists entertained for a while. Very little chance of direct technologicalapplications for the foreseeable future, though.

The team believes they have made the world's first room temperature superconductor, and they submit the paper to APL Materials (Impact Factor 6.6)?

Yes, there exist diamagnetism-paramagnetism phase transitions.

A room temperature superconductor would, by definition, maintain quantum coherence (of the superconducting order parameter) at room temperature.