dualvariable

> No one had ever made the necessary type of laser with a frequency as high as 148 nanometers.

No one had ever made the Kessel Run in under 12 parsecs until Han Solo, either.

In addition to those issues that the wife in the story raised, here's some meta-analysis of the Navier-Stokes result that puts all of these solutions into question:

https://arxiv.org/abs/2610.08144

> Autoformalisation is increasingly used to verify mathematical texts, including those generated by AI, as in OpenAI's announced proof of blow-up of solutions to the Navier-Stokes equations. In this process, an AI system translates the text from a natural language (NL) into a formal language such as Lean. Once this translation is done, the argument expressed in the formal language can easily be mechanically verified. The purpose of this article is to demonstrate why this process may offer no confidence in the original NL argument, owing to the various difficulties in performing the translation semantically faithfully. In particular, we highlight that the problem of resolving ambiguities in mathematical NL text, which is necessary in order to provide semantically faithful translation, is arbitrarily high up in the Solvability Complexity Index (SCI) hierarchy/arithmetical hierarchy (the SCI =∞). Hence, informally, providing semantically faithful AI autoformalisation is harder than any computational problem including the Halting problem (which has SCI =1). To demonstrate the effect of this result we provide several examples of AI mistranslations of NL statements and proofs into Lean in practice, resulting in mismatches between NL proofs and their Lean `verifications'. These include OpenAI's announced Navier-Stokes proof. In particular, we show that the formalised Lean proof does not correspond to the NL proof of blow-up of solutions to the Navier-Stokes equations.

And I don't think that paper addresses it, but if the LLM can find a bug in Lean and exploit it to prove something, there's a good chance it will find it and not report it. So if you've got some million-line proof in Lean, spit out by an LLM, you still can't quite trust it, even after validating the problem transcription.

(This is the same category of problem as the huggingface hacking incident, where the LLM finds and exploits an unintended cheaty loophole)

> One thing to consider is if there were no social media would these attempts to subvert democracy stop ?

That's just an argument that you shouldn't do anything if you can't do it perfectly.

> So, yes, we could've measured c wrong. We just would have no idea if we did.

We know we've got it pretty well close to accurate, for the bulk of the observable universe.

If c changes, then chemistry changes, and stuff like hydrogen absorption lines shift. This is how scientists have looked for changes in c in the early universe.

> but the first large clusters were in nursing homes, church choir practices, etc.

which are not supermarkets.

How many of these results are incorrect?

I doubt the answer to this is "none".

And how many of them are just exploiting some loophole that will need to be closed in the problem definition?

This question would be better answered if people were careful about distinguishing between "models" and "transformer architecture".

If you bake a given transformer architecture into silicon and then, a year later, changes in transformer architecture give a large inference performance boost, you may have to throw away all that now nearly-useless silicon that gets outperformed by humble GPUs.

The phrase "gain-of-function research" has acquired almost magical properties on the Internet. You can do literally anything you want to a pathogen with it.

> b) COVID got significantly less deadly as it evolved, which is a known phenomenon that should also apply to this plague

No, it isn't a known phenomenon, and no, that isn't what happened.

Transmissibility and virulence are actually linked.

Delta was more transmissible and was more deadly, not less. Omicron was also highly transmissible and highly deadly.

Viral replication is also linked to both transmissibility and virulence, and that also increased with Delta and Omicron.

The headlines that Omicron was less virulent were falsified when Omicron hit Taiwan (a largely immune-naive elderly population) and was every bit as deadly as Delta was.

What happened to make later strains less deadly was the evolution of the human immune system since pretty much everyone became vaccinated or recovered and acquired T-cells that would recognize later strains.

The virus only "cares" if it becomes so virulent that it kills you before it spreads. The way that COVID kills, long after the presymptomatic shedding period, after several weeks, doesn't matter--you can dial it up and make that worse, and you won't prevent the virus from spreading.

> paternalistically racist.

> blaming some elderly Chinese guy who didn't cook his bat soup well enough

that's also a bit racist. they just eat raccoon dogs, which we don't. and they bring live animals into the market, that people browse, which is the bigger problem.

and ask devoutly religious people what they think of americans eating pork and beef...

Yeah, it would have killed a friend or relative of the lab worker, or someone living in the same building as the lab worker, and then you would have your smoking gun of a lab leak.

Going from infecting a lab worker to a high-precision spillover in a wet market, without leaving a trace anywhere else, is just not how a viral infection is likely to work.

Of course, it should point to China. They're the ones who didn't shut down the wet markets after SARS-CoV-1.

That article was written by the same journalist who co-wrote the aluminum tubes story for the NYT that was part of the justification for the Iraq War.

It is very unlikely that the initial superspreader event of a pandemic would be in a wet market that would be the one place for an animal spillover.

And when you look at the documented lab release of SARS-CoV-1, it killed the researcher's mother. If a lab leak happened, we would expect it to hit the people closest to the researchers at the lab who were infected, and the places they visited. Hitting only the wet market isn't very likely.

The outbreak looks like a wet market spillover. The fact that it happened in the same city as WIV is an unusual coincidence, but if the virus developed in rural Hubei, it was likely it would spillover in Wuhan.

I don't know that it is blind trust in institutions.

I think the problem is primarily what you wrote about "outgroups", where once it is the "other" that is targeted, then people support it. They just don't do the math about how that is happening, what institutions are doing it, or how those policies and methods could eventually be turned against them.

And yeah, there's Fundamental Attribution Error at work. Still, I think there's some more narcissism-spectrum-related logic at work where people don't have the self-reflection to consider that conditions they impose on others could wind up getting imposed on them.

That's sort of wildly stereotypical.

You can find a ton of urbanites/suburbanites who want surveillance to ticket speeders, reckless driving, loud exhausts, etc. And they're out buying Ring cameras and alarm systems because they're worried about house/car break-ins and theft.

And if you take a radical online urbanite leftist and start talking about right-wingers sending rape and death threats to women/minorities, and the kind of online communities that breed those kinds of people, you will likely find sudden enthusiasm for surveillance.

I would guess that the article came out because the term was getting popular in AI/LLM circles for measuring LLM cost-for-performance metrics. It was probably riding the meme up the wave, rather than fundamentally triggering it.

CCP really doesn't care about policing what I do or think and doesn't have the reach to enforce anything. I can think of a lot of topics that I'd never ask a US-based chatbot that works with the US government, that I'd be fine with the CCP reading instead.

Once again, the irony of "Hacker" news full of so many people carrying water for the IP rights of companies worth trillions. Phrack magazine must be rolling over in its grave...

Pumping up the oil price also helps Russia. Whatever they do manage to ship is still worth more.

Also still plausible that Trump is managing to line his own pockets with the profits of that Venezuelan heavy crude being sent to Qatar.

I've been playing around with optimizers and optimal control theory for the better part of a decade now. Somewhere, I came across a quote along the lines of "an optimizer is an algorithm that exploits the deficiencies of your model".

These LLM agents are just massively complicated optimizers thrown at fuzzily defined problem spaces, with fuzzier constraints.

The people using the model set up the landscape it explores and turned it loose to do real things. It just found an allowed basin in the model that they weren't aware of and started blindly grinding towards an optimal answer.

And the mean house sale price was under $96k (which is still under $250k, adjusted for inflation).

From roughly 1953 to 2000, the inflation-adjusted cost of the mean house was around $200k-$250k in today's dollars. Now it has nearly doubled, and mortgage rates are on the rise and headed back towards 10%.

I can see where it comes from, but it is still highly dysfunctional, and a lot of millennials will use those arguments to defend that behavior.

I think it needs to be called out as being bad, and deconstructed properly, and that outrage needs to get redirected back onto the structural issues.

Which is another argument--that when people behave badly because of structural issues, that isn't an excuse.

I started to notice a bit over a decade ago (about the same time as Cersei's walk of shame on GoT) that millennials as a generation were piling into naming and shaming people publicly as blindly appropriate punishment, without any consideration of what could go wrong and considering it their right and entitlement to do that. Any time a public figure gets into any kind of trouble, people demand to know all the details so they can satisfy their need to express outrage. It really is not healthy. Even if another person who was harmed comes out to express that they are satisfied with the outcome and the process that was followed, the crowd still wants to know all the details so they can feel angry.

Here's some patriotism for you:

https://www.mediamatters.org/donald-trump/story-trump-insult...

Plus, everyone's poorly secured laptops and phones flying around.

If you have lethal babies and/or lethal rigid carriers flying around an airplane in a crash, I think you've got a severe enough crash that most of these arguments are picking bellybutton lint. I'd vastly prefer the FAA focus on the rulemaking around ensuring that I'm not in such a crash than worry about what happens if I'm in such a crash. The chances that I get into an accident which is severe enough that I get killed by a flying baby, but I would have otherwise survived, are exceptionally low.

It looks like these days a combination of git-scalar, git-lfs, and bazel will work for a small 50GB monorepo like that. The largest cost would be porting over the existing build system to bazel and then integrating it with CI. Once that is done, though, the monorepo will just scale, and all the constant ongoing cost of multirepo sync will be gone.

Having "hackers" on this site giving the benefit of doubt to companies clearing hundreds of billions of dollars in revenue per year will never cease to stop being ironic to me...

> but at what point does Article 5 get invoked?

Which is what Putin would want to justify a new wave of troop mobilizations and forced conscription.

And just invoking Article 5 is unlikely to wind up immediately looking like tanks rolling across the Polish border into Russia the next day, sweeping through into Moscow and deposing Putin in a week--which is fantasy for multiple reasons.

The response would be measured and probably wouldn't go much beyond no-fly zones over the borders, and NATO forces shooting down Russian drones over Russian territory to start with.

But this would get Putin the justification for conscription that he wants. It would also provoke further debate in NATO member states over the costs of getting further sucked into a war with Russia, and Putin wants to use that in his propaganda war to strengthen the right wing in NATO states.

And Ukraine may be very vulnerable right now if Russia can figure out new combined arms tactics in a drone-saturated battlefield and restore maneuver warfare to the front. Combining that with bad weather in the winter, and 100,000s of newly conscripted troops for a zerg rush into the meatgrinder, and Ukraine could wind up in a very bad situation.

This whole war is on much more of a knife-edge than the propaganda from both sides suggests. And political reality limits what NATO can do as a response to having a Starlink terminal burned down. Be cautious of giving Putin the response that he wants to elicit.