I kind of wonder if the best anti-distillation strategy is, if you detect suspicious traffic from someone (which happens a lot, they monitor for anything that looks like distillation), instead of blocking them, feed them say the output from Llama 3.1 8B or whatnot ;) Maybe finetune it a bit so it talks Claude-ish. But basically, subtly poison their dataset with hallucinations and crappy reasoning without it being immediately visibly obvious.

As for copyvio, sorry, this is something for the courts, and so far, the courts have not largely found against the trainers, and have instead found, by and large, that they're being compliant. The most notable setback against Anthropic for example was a finding that they couldn't just download books in training dataset off the internet, but that they could perfectly legally just buy surplus books for pennies on the dollar by the palletfull, scan them in, and train on that. That this is perfectly complaint with US copyright law.

I think a lot of you wish that copyright law was a lot more restrictive than it actually is. Which is a REALLY bizarre thing to see on Slashdot of all places, which back in the day was the beating heart of "Data Wants To Be Free!" philosophy.

To be clear, though... I would welcome a compromise modification to copyright law, which is, if you want to train on the public commons, you absolutely may, indeed, train on whatever you want, zero liability, but then you have to give back to the public commons. So maybe your top frontier models are closed, but you have to simultaneously release smaller distilled equivalent versions of it into the public domain (how to define "smaller distilled equivalent versions" is of course something that would require discussion), and release said frontier models to the public domain within e.g. 1 year or whatnot.

* They remain incentivized to keep pushing the frontier, since some people will always pay for the best
* They get permanently out of the worry of any copyvio liability (beyond basic requirements about not verbatim reproducing copyrighted materials in outputs)
* The public gets a constant stream of ever-better models, at no cost.

Sounds like a balance to me.

It doesnt think

Yeah, it does.

it doesnt rationalize

Yeah, it does.

These are not Markov chains. They're neural nets. They work via extremely complex chained fuzzy logic on superpositions of conceptual states.

And the less slop it has to deal with

This is literally a thread about distillation, aka, training on the outputs of other models. Synthetic data is the cornerstone of modern training. "Model collapse" is not something that actually happens in the real world, only in contrived settings, the model equivalent of if you could lock a person alone in a dark room with only their thoughts for ten thousand years.

What is "it" in this sentence?

Are you talking about the minuscule Google RAG model whose weights are probably measured in megabytes to lower-gigabytes and which you could run on a cell phone with capacity to spare, and comparing it to the multi-terabyte frontier models in question?

at least the person downloading such a component is responsible for "looking over it" before using it.

I'm sorry, but since when? You absolutely should not be tasking every single user of major projects with the responsibility of doing a code review on the whole thing before installing it. There needs to be a clear distinction between "fly-by-night thing that some rando uploaded" and "library that a million people depend on". In the traditional approach, the latter was something tat you installed with a package manner, while the former was something you went and fetch off e.g. Github or whatnot - and if you did the latter, then you were accepting that it was untrusted software. But now we have both (including the potential for "soundalikes") installed by the same means with no distinctions made by the install method. That is not good.

The funny thing about all this is that their only goal appears to have been to download publicly-available data off of UK government websites. I'm betting that they were (A) being tested on a knowledge-related task, and (B) not given direct access to the internet, but their tool capabilities involved access to the "gem" command (so that they could install Ruby packages for their work), and so decided to abuse (B) to cheat on (A).

Once again: training a model with the reward being "does it solve the task?" without looking at how it solves the task is very, very dangerous. This is basically the plot of Universal Paperclips.

I know a LOT of software that could use that as a pull request...

Of course the third bomb probably wouldn't have gotten used regardless of the order, as the next bomb (made from the infamous Demon Core) was scheduled to be ready to drop on the 17th, whereas Hirohito announced the surrender on the 15th (wasn't signed until 2 September, but it's hard to imagine that the US would have dropped the bomb after the announced surrender).

One of the amazing things is that the atomic bombings didn't shift the vote in the 3:3 deadlocked Japanese war council at all; the hawks (Anami, Umezu, Toyoda) remained focused on adaptation, not surrender. But it did make the doves (Suzuki, Togo, Yonai) more desperate, and Suzuki met with Hirohito. Hirohito had already been looking for an opportunity to surrender, having had plans to send Prince Konoe to Moscow to ask Stalin to function as a neutral broker to end the war (with secret instructions to accept "peace at any price"), but the trip was delayed by the Russian side, ostensibly by the Potsdam Conference (but in reality, to prepare for the invasion of Manchuria). When the terms of the Potsdam conference came out, Togo told Hirohito that they "were the most reasonable to be expected in the circumstances", to which Hirohito replied, "I agree. In principle they are acceptable." But the urgency to accept that in July wasn't present. Everything started collapsing in August, with the loss of their mainland and Pacific territories, the Soviet invasion, and the atomic bombings. What the bombings achieved was to help stir Hirohito to imminent action (which risked failing, as there was a coup attempt against him (the Kyujo Incident) to stop Japan from surrendering). The hawks saw unconditional surrender not as agreement to the loss of their sovereignty, but as agreement to enslavement.

But this is getting rather off topic. :) Main topic: if you think your AI is as much of a risk to the fate of humanity as the atomic bomb, which many of these people truly believe, then you can expect similar hesitancy as with the developers of the atomic bomb. They felt an urgent need - ending a brutal war - but at the same time had deep apprehension about what they were doing and how the fruits of their labour would be used.

And if you want a past example of scientists who feared their creation wanting to slow down, stop, or even undo it, BTW, let us consider the atomic bomb. Farrington Daniels took a poll of the Chicago nuclear scientists in 1945 asking what they wanted to happen with the atomic bomb. Remember that these were the people whose literal job was to make a weapon. Only 15% - 1 in ~7 - wanted the US to use it as it actually did, e.g. just start bombing cities just a couple days apart. Only a small majority wanted it even used in Japan, even against military targets without the Japanese first being invited to a demonstration in the US and being given ample opportunity to surrender. 13% - nearly as many as those who wanted the US to use it as it actually did - didn't want it used at all under any circumstance, in a war that had already caused a million American casualties.

The problem that they faced was that people like Groves** really wanted to bomb cities, and they, as the rank and file, didn't get a vote.

** - It seems dubious that Truman really understood what he was signing onto. While he never shirked from responsibility for authorizing it, the paper trail shows that he didn't seem to understand what he was approving at the time. In his diary before the attacks, he writes about the weapon, but about how he insisted it not be used against civilians, killing women and children, because he didn't want to sink to what he saw as the barbarism of the Japanese. In his first speech after the bombing, he refers to Hiroshima as "a military base", not a city, and says it was chosen to, insofar as possible, avoid the killing of civilians - but that if Japan did not surrender, then the US would bomb Japan's war industry, and then thousands of civilian lives would then be lost. Earlier drafts showed that it didn't contain the "insofar as possible" caveat. He genuinely didn't seem to realize what he was authorizing at the time. But Groves absolutely knew what he was doing, and deliberately selected sites for maximum civilian casualties. Truman seems to have come to understand what was going on by shortly after Nagasaki, where he gave the order to cease use of atomic weaponry, saying that the thought of wiping out another 100,000 people was too horrible and that he couldn't stomach the thought of "killing all those kids".

They are competitive in quality with American models.

They absolutely compete against lower-end or previous-generation US models. I use GLM-5.3 and GLM-5.3 flash a ton, for example. But they do not compete against the top-end US frontier models from Anthropic and OpenAI.

If you can do a normal difficulty, verifiable task with an open Chinese model, you probably should. GLM is less than $5 per 1M tokens out, vs. $50/1M for OpenAI Astra for example. But there are some things that you'll just get stuck grinding your gears on with the former that the latter will have no problems solving. Or it might be a really important nonverifiable task, e.g., not "does this program work".

Also, I'm not sure what "defenses against distillation" is supposed to even mean.

I've never met a Silicon Valley company that decided to slow down for reasons that weren't also aligned with business interests.

These people genuinely are scared of their creations.

I'm hopeful for you :)

Basically, the regulatory mechanism is: NO is "the brakes" on the mitochondria, and it's also a vasodilator. More NO in a tissue is a signal that it needs more oxygen, and it "opens the floodgates". The problem is that the rate of reaction in the mitochondria is proportional to the local oxygen concentration, so the cells nearest the supply will tend to disproportionately consume it. So the electron transport chain in these cells are simultaneously "braked" by the NO to limit how fast they can consume oxygen.

But the NO-CcO balance can get into a pathogenic state of excess NO (nitrosative stress), where the electron transport chain basically grinds to a halt, leading to high amounts of electron leakage and the formation of lots of superoxide. This combines with free NO to make peroxynitrite, which does irreversible damage to the mitochondria and will eventually kill them (this can cause a vicious cycle, because as the cells struggle to get enough ATP, they wrongly interpret it as hypoxia (pseudohypoxia) and trigger the release of more NO to bring in more oxygen. Even worse, as the failing mitochondria undergo apoptosis or necrosis, mtDNA leaks out, which the cell wrongly interprets as a bacterium, triggering the release of pro-inflammatory regulators, which trigger the release of... wait for it... lots more NO.

Nitrosative stress has been traced as a pathogenic mechanism in a lot of diseases, everything from diabetes to parkinson's to athlerosclerosis to ALS to asthma to cancers to IBS and on and on. So any new mechanism to help clear nitrosative stress, even if just temporarily, could help with symptom relief from a really broad range of conditions.

"Humans have an impulse for action and AIs don't" is your response to an article about an AI grinding its gears, ala "NEW REALIZATION -- I'm burning a lot of time on hCaptcha round-trips .... so the answer payload shape is right, the token+image pairing is right (from the same script.js!), cookies are right (requests) and STILL 'wrong answer'. SO WHAT THE HELL IS WRONG WITH THE ANSWERS?" "?

I'd argue if anything, it's humans whose drive for action is weak.

It is if the damage doesn't progress to necrosis.

I never timed its disappearance when it happened to me (I'm not the GP), but I'd day maybe a couple weeks? Not really something I gave much thought to.

Also, it's perfectly possible that the red light from being outdoors in the sun is also good for you - but that also comes with other frequencies, including harmful ionizing UV light, which "ages" the skin and risks skin cancer (though the increased vitamin D from ~290-300nm is an offsetting factor).

Red light therapy is one of those things that sounds like total woo hokum when you first hear about it, but that studies keep showing benefits to.

Key thing to remember is that these studies are using very high doses of red light - often limited by the amount of heating you can safely tolerate in the subject (not heat from the emitter, but from the photons themselves).

Honestly, it doesn't sound crazy at all. Different tissues have different light susceptibilities (that's how, say, laser hair removal works), due to the susceptibility of specific compounds. Red penetrates very deeply (remember the classic trick of shining a flashlight through your hand; only the red makes it through). Red light has a number of very specific interactions it causes. The main one is that it photolyzes the bond between NO and cytochrome c oxidase (triggering NO signaling, causing vasodilation). A secondary possible affect is that it affects light-gated ion channels (esp. in the TRP family).

And the amount of red light we're talking about is very high. Instantaneous fluxes are like 1000-5000 times higher than ambient indoor light in the red spectrum, and even averaged across a day it's still like an order of magnitude higher than a person who spends their day indoors. For NIR therapy, the difference is even more extreme (commercial LEDs emit essentially zero NIR). While outdoor light exposure is much higher - 15 mW/cm2 red + 20 mW/cm2 NIR, comparable to the 10-100 mW/cm2 of red light therapy - it's broad spectrum, while red light therapy is concentrated around cytochrome c oxidase peaks, so the flux around these specific peaks is ~10-50x higher with red light therapy.

Flux matters a lot because to meaningfully dissociate NO from CCO, you have to do so at faster than the rate that NO and CCO recombine. Sunlight on its own (let alone indoor ambient light) don't do that. It doesn't matter if you're spending all day in the sun vs. 15 minutes in treatment if the sun isn't hitting the threshold effect level.

Another difference, BTW, is not just "shooting it up your nostrils", it's scattering. Sun just hits your skin and much of it is lost through scattering and absorption in blood, esp. hemaglobin. Red light therapy targeting deep structures is delivered by a contact probe that compresses the skin, displacing blood. This increases penetration into deep tissues 2-5x.