There are sequential dependencies, so you can't just arbitrarily increase speed by parallelizing over more GPUs. Every token depends on all previous tokens, every layer depends on all previous layers. You can arbitrarily slow a model down by using fewer, slower GPUs (or none at all), though.
Anthropic is obviously also aware of the benefits of MoE and distilling a larger model into a smaller one, so they could run a model of the same size as Alibaba's for the same inference cost if they want to. Or they can run a slightly larger model for slightly higher cost. They definitely aren't running a much larger model (except potentially as a teacher for distillation training) because then they wouldn't be able to hit the output speeds they're hitting.
> The new 1s have a mandate to at least run on local hardware.
They do? Source?
But if that's true, it would explain why Minimax, Z.ai and Moonshot are all organized as Singaporean holding companies, with claimed data center locations (according to OpenRouter) in the US or Singapore and only the devs in China. Can't be forced to use inferior local hardware if you're just a body shop for a "foreign" AI company. ;)
Copyright protects even very abstract aspects of human creative expression, not just the specific form in which it is originally expressed. If you translate a book into another language, or turn it into a silent movie, none of the actual text may survive, but the story itself remains covered by the original copyright.
So when you clone the behavior of a program like chardet without referencing the original source code except by executing it to make sure your clone produces exactly the same output, you may still be infringing its copyright if that output reflects creative choices made in the design of chardet that aren't fully determined by the functional purpose of the program.
Yes, of course there are people in China who, when their job puts them in conflict with their ethics, will decide to do something ethical. I can't think of any war-related examples, since it's been a while since China was involved in any big wars, but I like the story of Liu Lipeng, who used to work as an internet censor: https://madeinchinajournal.com/2025/04/03/me-and-my-censor/
Those aren't even percentage increases, but standardized effect sizes. So if you take an individual survey respondent and all you know is that they self-reported higher AI usage, you can guess their answers to the self-reported individual effectiveness slightly more accurately, but most of the variation will be due to unrelated factors.
The question that people are actually interested in, "After adopting this specific AI tool, will there be a noticeable impact on measures we care about?" is not addressed by this model at all, since they do not compare individual respondents' answers over time, nor is there any attempt to establish causality.
People could always simply argue in court that their torrenting was free use.
If you're just some nobody representing yourself instead of an expensive lawyer acting on behalf of a large company, maybe the judge will even try to be extra nice when he explains why the argument doesn't hold water.
Sadly, in many courts, when it comes to the corporate and the government, the judges rule on the axiom, "Show me your lawyer first, and I will rule, rather than show me the law, and I will rule".
> maybe the judge will even try to be extra nice when he explains why the argument doesn't hold water.
The thing everybody ignores about this is context.
Suppose you upload a copy of a work to someone else over the internet for <specific reason>. Is it fair use? That has to depend on the reason, doesn't it? Aren't there going to be some reasons for which the answer is yes?
The "problem" here is that the reason typically belongs to the person downloading it. Suppose you're willing to upload a copy to anyone who has a bona fide legitimate fair use reason. Someone comes along, tells you that they have such a reason and you upload a copy to them. If they actually did, did you do anything wrong? What did you do that you shouldn't have done? How is this legitimate fair use copy supposed to be made if not like this?
But then suppose that they lied to you and had some different purpose that wasn't fair use. Is it you or them who has done something wrong? From your perspective the two cases are indistinguishable, so then doesn't it have to be them? On top of that, they're the one actually making the copy -- it gets written to persistent storage on their device, not yours.
It seems like the only reason people want to argue that it's the uploader and not the non-fair-use downloader who is doing something wrong is some combination of "downloading is harder to detect" and that then the downloader who actually had a fair use purpose would be able to present it and the plaintiffs don't like that because it's not compatible with their scattershot enforcement methods.
> It seems like the only reason people want to argue that it's the uploader
Well there's also the issue of enablement. If you're overly enthusiastic to turn a blind eye to illegal conduct you end up being labeled an accomplice. But of course that would seem to apply to Facebook here in equal measure.
> Well there's also the issue of enablement. If you're overly enthusiastic to turn a blind eye to illegal conduct you end up being labeled an accomplice.
That's something the industry made up out of whole cloth. If someone sells ski masks that can be used for both keeping the wind off your face when you're skiing and hiding your face when you're committing burglary and has no means to know what any given person intends to do with it, are you really proposing to charge the department store as an accomplice?
The way that would ordinarily work is that you could charge them if they were e.g. advertising their masks as useful for burglary. But now where are we with someone who doesn't do that?
> That's something the industry made up out of whole cloth.
No, that's how the law (at least in the US) generally works across the board.
Your ski mask example is misplaced. There are legitimate uses for the product and it isn't immediately apparent to the store why someone might be purchasing a given item. It's not their job to invade their customer's privacy.
Your logic regarding torrents only works if we assume that a significant number of peers are engaging in fair use. For a torrent containing copyrighted content that was never distributed by the rights holder via P2P. Thousands of peers all working together to make backup copies of their legitimately purchased products. Right.
I'll freely admit that scenario to be beyond absurd despite being no fan of the current copyright regime in the west.
> There are legitimate uses for the product and it isn't immediately apparent to the store why someone might be purchasing a given item. It's not their job to invade their customer's privacy.
You seem to be implying that nothing is ever legitimately fair use.
> Your logic regarding torrents only works if we assume that a significant number of peers are engaging in fair use.
What evidence do you have that this isn't the case? Not evidence that someone is doing it for the wrong reasons, evidence that no significant number of people are doing it for any legitimate reason.
There are a lot of things that are plausibly fair use. If you subscribe to a streaming service with a plan that includes 4k content but their broken service won't play that content on your 4k TV, can you get a 4k copy of the thing you're actually paying for? If you're a teacher and the law specifically says you can make copies for classroom use, but the copy you have has copy protection, can you get a copy that doesn't from someone else? If you're an organization trying to make software that can automatically subtitle videos based on the audio so that hearing impaired users can know what's being said, and you need a large amount of training data (i.e. existing video that has already been subtitled) to do machine learning, can you get them from someone else? What if you're an archivist and you actually are making backup copies of everything you can get your hands on to preserve the historical record?
What alternative do such people even have for doing the legitimate thing they ostensibly have a right to do?
Judges often roll this line out, but in criminal court I've seen some defendants get epic deals by going without a lawyer [0] since absolutely nobody in the justice system wants to deal with the guy who has no idea what he's doing and is going to make the most bizarre arguments about being a sovereign citizen. So they give them a really low offer and get them on their way as quickly as possible.
[0] I don't like to say "represent yourself." I once angered a judge by pointing out that you can't "represent yourself, you are yourself."
There's lots of wiggle room if you know how to start with the proper standing. The problem, from my perspective is: if you are trying to get away with wrong doing, or if you are legitimately being harmed by the law as it's being applied. If you are asking questions or making statements. Everything the court says is a legal determination, everything you say can be used against you,.
Im sure a lot has changed no with much criminality in the justice system and government.
No, but it does matter how much money the alleged infringer has.
Property law is mostly concerned with protecting the rich from the poor, so when a rich person violates the property of a poor person, the courts can't allow the inversion of purpose and will create something called a "legal fiction," which is basically the kind of bending-over-backwards that my children do to try to claim that they didn't break the rules, actually, and if you look at it in a certain way they were actually following the rules, actually.
Yes, the VC-backed startup ecosystem that was the origin of this website does rely on propagating the myth that we live in a meritocracy to ensure it has enough cheap labor to build prototypes that its anointed few can acquire at rock bottom pricing. But we've been through enough cycles of it now that we've started seeing the patterns.
History clearly establishes that the open market assigns substantial value to human life. We just happen to have outlawed trading in it. Human life has been deemed worthless by force of law.
Less facetiously, you're committing a semantic error.
"Markets clear" is one of those meritocracy myths that we the hoi paloi get taught explicitly all the while the elite will tell you to your face they don't believe. Google and Meta are massively profitable companies built on the idea that the concept of value is manipulable.
maybe the judge will even try to be extra nice when he explains why the argument doesn't hold water.
Many judges take a dim view of expensive lawyers trying to pull the wool over their eyes with sophisticated but fallacious arguments. You have to deal with a lot of BS to be a long-standing judge, so it seems like resistance to BS may be selected for among judges.
You mean the fee I pay for piracy doesn't cover the cost of the piracy? Maybe they should remove the fee, so they can prosecute me for piracy, without me arguing it's covered by the fee.
reply