Imagine having the brass neck to go and lobby for a rent-seeking monopoly?
AI labs need to innovate if they want to win. Open models are capitalism. Ideally Western companies should get more efficient, but if we can't then we should be doing the opposite of regulation, and subsidizing western models.
OpenAI and Anthropic kind of need that revenue to justify their current valuations. The US market by itself is too small for that. And if that money dries up, it's going to be much harder for them to maintain their technical head start. The US starting to impose all sorts of restrictions here would not be helpful.
That's funny, because they made their models exactly because of open stuff on the web.
But they will loose this battle. There is no way to technically avoid open models to be created, so they will be created.
Now they’re being outcompeted and so we should suddenly have regulations.
Time to put your big boy pants on and compete.
This is a Big Green Flag for open-weight AI models.
The OpenAI/Huggingface debacle is patient-zero. Huggingface used GLM 5.2 to help mitigate and were unable to use more capable closed sota models for the analysis due to guardrails. It is a foot gun of the highest order. We should all be concerned. I feel like Captain Obvious saying this.
Thank goodness no one is taking any of this seriously, because it could never be anything more than a machine generated hatchet job.
A faster and better educated understanding of the subject matter could be achieved by standing in front of a wood chipper and dropping a brick in it to see what happens. “Yes, there were results! But why? Why to literally every variable involved?”
They act like people don’t see it and then wonder why there might be any resentment towards AI companies.
I am pro AI but there has long been a cultural problem with mercenary attitudes in the big data/ML and now AI world of which this is just the latest most extreme manifestation.
Make laws, make policy, ban them from US markets. It won't stop China doing whatever the hell they want. Like SpaceX developed reusable rockets and China thought "That's a good idea" and now they have their own reusable rockets.
They will do that with AI too, and fast. Having your head in the sand just exposes your rear end, and you've removed your ability to see the predator coming...
I knew this would happen someday.
So it isn’t just the two companies and their investors that have a vested interest in keeping the music playing, it’s the entire US government. No surprises then that government officials are now openly accusing Chinese companies of cheating. Sanctions and tariffs are probably going to follow.
What happens when they help authoritarian dictatorships build nuclear weapons?
What happens when they teach average people the steps needed to make a bioweapon at home?
Every society restricts dangerous capabilities at the end of the day, because most people realize there is a crossover point between the danger of the capability and it being available to everyone.
It's a principled take that seems naive from a certain point of view.
Some more very bad people are building them.
This democratizes that ability.
Might be good. Might be bad. Might be both.
Fuck you for thinking you know better than everyone else.
The reason barking at open source will only get the dog thrown into the river is because there is a whole lot more cumulative power distributed among companies and consumers who don't want two incumbents to dictate which kinds of alternatives can exist.
Note that those rushing to the "all corpos evil" talking points will turn around and advocate delegating these decisions to "the government." This is the same government that they also claim is bought by the corporations. It's also the same government that will enforce the regulatory capture on behalf of the yucky corporation against the broad, diverse interests.
Thinking big vs small while embracing the biggest of the big is some kind of inconsistency that people need to deal with by looking in the mirror and finding some awareness. Nuance. If your problem model is wrong, so is every solution you propose except if your wrong model may get lucky from time to time. Not every broken model is a broken clock. Some broken models cannot admit the right answer ever.
The real dynamic on display here is broad, correlated mass of overlapping interests, big and small. The big pile of people and entities that don't want to be governed by OpenAI and Anthropic can push harder and win the tug of war. You don't have to like your temporary allies all the time, but at least know who they are and when you are aligned.
If I wager on anything else, it won’t matter, because it would mean living in a dystopian hell where any bet is meaningless.
In recent years, with US further turning their back from globalization into America first protectionism, the rest of the world is getting increasingly tired and slowly de risking out of US products and tech.
European companies are either suffering us levies, or us interference with whom they can conduct business with, or consistently risking that their service and cloud providers can become liabilities over night.
EU is also strongly pushing for US alternatives in payments, telecoms, space, etc. And I didn't even mention politics yet.
US is still enjoying an apex position, but it may overestimate its gripe and underestimate resentments (Greenland threats have been a point of no return).
The actions of US looks like those of a decaying empire.
What plays in US hands is that its biggest "allies" are mature democracies, and mature democracies really procrastinate any painful change unless they have to confront it, thus it may keep getting away with this for very long.
One wonders how many times that trick can be used. In fact, no need to wonder. Look around, the EU is busily spinning up arms industries, alternative payment platforms and expanding trade agreements with nearly everyone else.
Yes. The US can try. But while we haven't filed for divorce yet, we are definitely looking elsewhere.
Fuck Sam and Dario.
These fucking assholes want to selfishly sink our country and our competitiveness because their valuations are tied to a thin little moat, and they moronically raised way too much capital. They think regulation can save them, when in actuality it will doom all of us.
Go to hell, OpenAI and Anthropic.
Call your senators right now.
The only practical open source models, that can be fine tuned and are of sufficient quality are all Chinese. The only thing that has been released by labs are yesterday's tech. And that's exactly how they want it. They don't care about open source, don't care about the community. It's just another funnel for them to capture customers into their closed tech, which is where the actual serious high quality models are.
There is this chronic refusal to believe that new technologies can cause harm.
My team works on testing risks from both companies' leading models and I feel pretty confident saying that uncensored next gen models becoming "open" would be a net negative for society.
I don't understand why no narrative regarding the risks of technology can land with the tech crowd until it's too late (hello, social media!)
Yes, I totally love the fact that when using the closed source models that I pay for, once released, they can still shut me out or silently poison the results based on IP address, timezone, hostname, keywords or just the general vibes, etc. at any time (all of which already proved to have happend or even admitted to / bragged about). /s
Yeah. Nah.
Bringing out the same drum I’ve been banging for years now, these companies have never actually cared about ensuring safe AI, regulation, or ensuring it benefits all humanity. It has always, always, been about control and their profit margins. Their actions have never reflected any sort of positive, shared human benefit since OpenAI released ChatGPT, and they’ll never actually provide said shared benefit when it comes down to benefitting everyone versus benefitting themselves.
The technology is built off wholesale theft of protected works without compensation; at least China gives most of those models back, for free, in the form of open-weights.
It’s the Princess Bride meme but for AI: “you’re trying to take what I’ve rightfully stolen!” I’m tired of having to put on some sort of mask of decency and professionalism to discuss a topic whose counterparty gets to wallow in toxic waste without recourse, and these sorts of stories only reinforce that despite being a glib jerk about it, I’m still on the high road, here.
More folks should join us in open mockery of hypocritical whinging from technological robber barons.
A quick scan shows a lack of rigor that usually points towards zealotry rather than a reasoned position: you imply that models are only trained on copyrighted work, you don’t address the ridiculous “copyright forever, nothing goes public domain” policies that got us to this place, you don’t distinguish between training on an 80 year old copyrighted work whose author is dead and rights are just in a company portfolio, and you assert that neither OpenAI nor Anthropic has ever done anything altruistic.
If you enjoy the screaming, have at it. If you want to persuade people, dialing it down a few notches and at least giving token acknowledgement that nuance exists would help.
If anything your “copyright forever” drumbeat and incorrect read of “only being trained on copyrighted works” are more hyperbolic than anything you’re responding to.
Also, the models from China are only being released open weights as competitive positioning against the globally more used competition. If the Chinese companies had a dominant market position they’d be falling prey to the same forces that have caused OpenAI and Anthropic to behave the way they do.
1. You're mischaracterizing and exaggerating the policies to falsely support your point. Things go into the public domain all the time: https://web.law.duke.edu/cspd/publicdomainday/2026/.
2. It's not like the model companies had the attitude "oh copyright is too long, we disagree on the term." Their attitude was precisely: "We want it and we don't give a fuck about you. Published yesterday, published 50 years ago? We will take it, we won't pay for it, and we will use it to replace you and make ourselves rich. Don't like it? Suck my data center."
You're describing what Pirate Party proponents have been championing since the start. Not that I disagree with that notion. Copyright law needs to be changed for modern times. But the laws as currently exist are what they are.
The question is whether people like Altman, Amodei, Musk are part of the common case in Hanlon's razor or they want to trigger emotional responses to gain agency in the mess.
You're doing the same thing:
> you imply that models are only trained on copyrighted work
Actually, they said "The technology is built off wholesale theft of protected works without compensation". The word "only" never appears here, nor is it really implied.
> you don’t address the ridiculous “copyright forever, nothing goes public domain” policies that got us to this place
This is pure "whataboutism". Both things can be bad.
I mean, you didn't address wealth inequality or global warning.
Not to mention that those nuances, as you state, mean very little. To say it was trained solely on copyrighted material or to say it was trained in large part on copyrighted material doesn't much move the needle over the fact it was done, a lot. Making a distinction between an "80 year old copyrighted work" or a recent one is also moot as it a matter of public record they themselves didn't make that distinction and happily fed on both kinds.
On to not being able to say they didn't ever do anything altruistic, well animal rights and welfare had great strides under the Nazi regime, not sure it evens out the rest. The same logic can be applied to the current state of affairs, doing a little good does not even out a lot of bad.
In closing, nuance is relevant when its relevant, after that its just being pedantic.
You’re not going sufficiently far back in comment histories (or using AI summaries of it from existing search tooling), or viewing the entire body of work I’ve written, shared, or discussed about all of those topics for the past twenty years.
I scream and wave my arms because nobody cared for decorum. I cut to the meat of the argument because nobody cares for context. Hell, this is the comments section for crying out loud; nobody is expecting a nuanced treatise about systemic interactions of multiple market and policy forces that resulted in this specific outcome in the comments section, even on HN.
Further, I suspect most folks here don’t actually care. I scream and wave my arms because it’s actually had the best results in outreach thus far - people see the flailing, read the bullet point, check the history, and somehow decide to reach out to me for more context (which I am always happy to give). When greeted with a wall of plaintext on an algorithmic site, the trained human response is to basically fuck off and go find a dopamine hit elsewhere, not stick around and read someone’s shoulda-been-a-blog-post on a social media comment page.
For all the crowing about “decorum” and “professionalism” and “intelligent debate”, the grim reality is that such concepts can actually detract from arguments by making it seem like we have time to discuss these further, that things aren’t that bad yet, that we can really think this through.
To which I say: no, we cannot. This language is a dire threat, it is a crimson red flag of intent to deliberately cause harm. You do not make an impassioned and polite argument about the nature of market forces or copyright policy with a mugger brandishing a gun, you do something about it.
For now, and into the future, I hope that we protect that unique and novel idea of an anthropomorphic mouse, which should be expanded to cover all forms of anthropomorphism, if we're being honest with ourselves.
And, to that point, the protection of that idea should also extend to the (clearly infringing) hit 90s YA book series Animorphs, as it somewhat sounds like anthropomorphism.
Protect their rights! Anything produced within the conceptual acreage that has been clearly marked and claimed is theft!
Tech companies, in the end of the day, will stand together to suppress dissent under the guise of anything nearing plausible deniability.
i was curious, because stego-tech also misrepresented the reason for their mod warning, and yours is the same.
you didnt get a warning because you were "discussing flock". its because you made a boring comment about "oligarch slushfund to pay bots" for "pro-flock astroturfing". that's not "discussing flock", its vague accusations of astroturfing.
its like, super duper clear in the guidelines that if you suspect astroturfing or botting that you should email the mods so they can investigate it.
dang has some pithy line about "mod less, not more when there is a relation to ycomb" and its true. ive ripped on flock plenty of times (fuck flock!). just put some actual substance in your comments.
openai and anthropic and google etc hire PHDs to create datasets by solving problems, writing thinking traces.
then they train their models on those datasets.
also as another user said, what is this 80+ years to protect some text? its ridiculous
Companies don’t have souls or minds. Some people in some large companies have absolutely cared about social good, art, and other things outside of financials.
However, a half of century ago there were a non-negligible number of large companies for which there existed more important priorities than the financials, e.g. reputation and customer satisfaction, for instance Hewlett-Packard (which was something extremely different from those who have inherited this name).
I am with you 100% but mockery will not affect their wallets. which is the only thing they care about.
You’re more likely to laugh at a comedian when everyone else does, even if you don’t get the joke.
It was never the intent. What we are seeing is the methodology that Peter Thiel (hello, ycombinator) described in his book Zero to One [0]: "Competition is for losers"
Preach! Anthropic and OpenAI care about one thing, and it's not safety or the benefit of humanity: it's regulatory capture. All that other shit they wring their hands about – Anthropic especially – is invariably a marketing tactic or a persuasion tactic meant to convince us that A) they have the most capable, dangerous AIs and wouldn't you like to use them; and B) only they can be trusted to develop them because look at all these precautions they take and all of the blog posts/tweets their founders write about the oncoming doom of humanity.
Being an "AI inference middleman" is pretty lucrative, and AI companies have only started to crack down on it.
China also publishes a lot of impactful research in the open which has benefitted US labs too.
Memento mori.
https://techandmedialaw.com/open-patent-rights-for-seatbelts...
(edit: sorry, re-reading your comment you were more directly referring to the AI companies and this context, I interpreted this more generally on my first glance)
Look, if I was worried about nuclear bombs, I wouldn’t work on a nuclear bomb developing research center.
It’s all fake and pompous garbage.
Thinking that an actor in a system will go against the incentives of the system (at least at scale and over time) is naive.
It’s partly not arbitrarily punishing the naive unless they refuse to learn the lesson. Invite them in, acknowledging the error in judgement, and help them to be better going forward.
That’s how we win.
it wasn't really "be nicer" and im surprised that was your takeaway; your last comment which was flagged had no substance. it was boring. ("i hate american aitechbros". neat.) at least this comment has some conversation starter in it.
> Meanwhile America is tying itself in knots: politicians and bureaucrats are banning new data centers, piling on state regulations, and pushing for new federal agencies to pre-approve frontier models.
> This is how you lose the AI race. The rest of the world won’t play by our rules if we bog ourselves down. Permissionless innovation is how America won the internet and became the technological envy of the world. We can do it again with AI -- while addressing risks in a targeted way -- or we’ll watch our lead evaporate. [1]
- David Sacks, Chair of the President's Council of Advisors on Science and Technology
There are lots of Anthropic and OpenAI employees here.
There are also lots of people that totally hate AI here.
If they don't like your comment, some of them will flag you. I get caught up in this frequently. The flagging strongly correlates with the audience that the headline sentiment attracts.
In a thread like this that is laser-focused on real criticism of OpenAI and Anthropic you'll do just fine to call out their bad behavior.
And I'll take that 10 times out of 10 over China's role in all of this, which is specifically to disrupt and sabotage the rest.
I can trust Anthropic and OpenAI to have economic interests. I can't trust the same from Chinese companies, which I can actually trust to have geopolitic interests.
The only country with anything to gain or lose here is the United States because it decided to play the game of market domination. Either it ends up with market domination and wins, or it doesnt and loses. China's result will remain what its set out to do, which is to create a technology that will support millions of small bespoke companies unbeholden to a larger firm serving all sectors of the economy.
OpenAI has the year-old GPT-OSS, and Anthropic has never released any open weights models. Meanwhile, other companies are delivering a steady stream of high-quality open weights models every month, and I'm not going to stop using them just because some other people don't like the country they were developed in.
Anthropic especially has this arrogant, “we’re smarter than you”, communication style, and acts like they’re serving some higher purpose (to the point of absurdity in some cases like “model welfare”) - I think part of all this might be the EA roots but I’m not sure. But then they get a little competition and suddenly they’re a little tattle-tale running to the government looking to artificially preserve their advantage. I’ve seen nobody who doesn’t see the naked self interest. There’s no mission, not higher purpose, not even competitive spirit, just some people that were on top and can’t handle the idea that they weren’t as special as they thought. It’s going to bite them in the long run, we’ll probably get some politically motivated legislation, but the cats out of the bag, they’ve lost all credibility and don’t even have a uniquely competitive model anymore, they’ll end up as just another also-ran.
Their users? (Not to be confused with their users' money)
Winning strategy.
…
The U.S. should pass a law that (1) makes explicit that collecting data for training models is fair use, and (2) bars terms of service that forbid distillation, for U.S. companies at a minimum. Stopping distillation — which is literally just querying the API — is nearly impossible; the U.S. should go the other way and lean into a new copyright policy that both indemnifies the labs and also guarantees that what they learned fuels further innovation for everyone else” [1].
[1] https://stratechery.com/2026/whos-afraid-of-chinese-models/
Let it all burn:D
edit:
And lets hope we learn a lesson.
That is the last thing that will happen. We don't seem to ever learn a lesson.
Winning the AI race means letting OpenAI and Anthropic face market forces.
Edit: Clearly other companies are doing well. ElevenLabs, Runway, FAL, RunPod, CoreWeave, Lambda, BaseTen, Lovable. These are real companies that didn't raise absurd amounts of capital. They've grown revenues at a solid clip, provide real value, and didn't go off the deep end with fundraising.
This is a ultra speed run of what Red Hat and SuSE upending Windows in the 90s/00s.
LLMs are incredible and extremely useful, but they never should have been a for-profit venture. At least not at first. They should have originated from universities (who would have had the capital to fund model development and do so ethically in this universe), and LLMs should have been much, much smaller in scope (like autocorrect in word processors or autocomplete in IDEs). Instead, we (well, OpenAI mostly) opted to burn the Amazon to extract profit from the wood, our planet's future be damned.
Between this, transferring billions from EV R&D, eliminating the EV tax credit and targetting research universities, I can't help but think that Krebs was right. So MANY of the policy decisions being made by the US are inexplicable UNLESS you assume that a rootkit has infected the government.
Europe regulates to protect humanity.
USA regulates to protect monopolies and political power
So, I am not sure it is such a win. These guys seem to be winning.
You don't hand every random script kiddie a zero day button. When models get better, you don't hand them an AGI that can help them make bioweapons or whatever.
When it comes to intelligence, there is a limit to openness that any sane non-tech-libertarian would agree on.
The big picture: ...
Zoom in:...
Zoom out:..."
I wonder if they used an American or Chinese LLM to write this article?
Still, if feels very LLM-y with their short sentences, definitive statements and a love for lists when none is needed.
On the other hand, maybe Axios was such dominant component in the training dataset that now every LLM speaks like this...
People rightfully question if these business models were ever feasible or sustainable to begin with. But isn't that the risk they took? We can certainly recognize that being displaced before realizing profit from your work is like a kick in the pants, but isn't that always supposed to be a possible outcome? I realize that is part of why we have intellectual property rights to reward innovators - but that isn't a universal concept in this global economy.
Which, some might argue, may be the point on the global stage. It is an ideological competition as well.
They are part of tribal play to make everything rented. Self driving taxis, where no one owns a car, rented computers talking to rented compute for AI tokens, rented text books, and rented history books preaching hallucinated facts
That tiny cabal wants everyone to be digital serfs where the only place to buy or sell is on a platform owned by a tribe member while you ride in your rented self driving taxi to your rented house while making dinner plans using your rented compute.
China wants you to own your bits. Because a 30 year old tractor you own is better for a small farmer than a rented John Deer self driving auto farm in the hands of the serf.
The American tribal vision is serfdom forever and the Chinese vision is everyone having a tiny farm they own with local compute AI driven cars that cost half a years salary instead of three years salary with continuing rental feature subscriptions.
I know which side I am on.
Do your part for humanity and replicate Chinese open source models as far and wide as possible! It is for our future to avoid the Tribal's push for Serfdom! China is trying to help us!!
Use Chinese AI today. Its for your children tomorrow or they will be serfs to about ten Jewish men and one South African Satanist with a dozen Jewish babies.
I don't think this is the main thing. I think the real risk open weights AI poses to them is that it allows us to get inference from providers that only care about making profits on inference. What OpenAI/Anthropic want is to be able to jack up inference prices once everyone is dependent. Open weights keeps them honest.
This website just does not fundamentally believe we are getting to AGI, or that RSI is a thing that can happen, and so anyone doing anything about it must have some evil ulterior motive.
2) There are many reasons why Anthropic and Open Ai will collapse aside from this.
3) Wait until you have non-China models that are good enough. Hint: they already exist. Microsoft partnering with Mistral is a big red flag for OpenAI and Anthropic.
"Then you're not pro self defense!"
Why not? These tools are also very useful for defense, you just need to apply them.
Read the HuggingFace postmortem, they had to use an open model because all others they tried refused to do any work.
It's a gold rush, and the only guy getting rich will be the one selling the shovel (nvidia + cloud in this case).
It might mean we'll get cheaper cloud compute in the future, but I'm also positive on the market's ability to absorb the excess capacity quickly. It's Jervon's paradox.
https://www.quora.com/Why-did-Steve-Ballmer-call-Linux-a-can...
I guess it hurts the big AI companies knowing that AI can be way cheaper than what they sell.
As I said a day or two ago, the open weights Chinese models may be good (hey might be better even) but let's not fool ourselves - the only reason the Chinese are pushing for open weights is because they were not the first people at the party and that's the only way they can fight against openai and anthropic. They are just as much of an enemy as both openai and anthropic - hopefully enough people realize that without supporting either. Just sit back and let them eat each-other. That's the only sensible thing to do.
Certainly people have preferred tools. And if people argued with you about the merits of llms, that’s obviously a different thing.
I suspect if a third provider blew the two out of the water tomorrow with an amazing model, people would want to try it right away. Assuming the company wasn’t grinding up kittens to produce it, or something equally noxious.
RIP Anthropic "good guys" January 2021 - January 2026.
Earlier this year they limited how you could use your subscription. It's been rapidly downhill since then.
I shouldn't get in arguments about this stuff online, but have you actually sat down and thought about this in-depth? It's pretty normal to have a gut reaction that this is insane, but it's worth thinking about.
If we replaced every neuron in your brain, one by one, with a silicon equivalent, would you stop being conscious? At what point? Would you be able to tell?
If yes, what about if we took those silicon neurons and instead simulated them, IE perfectly simulated an entire human brain, neuron-for-neuron, would that be conscious?
If yes, where is the line at which code starts being conscious? Can you categorically say that existing LLMs are on one side or the other? Would scaling up or down change that?
There are robust philosophical positions on all sides of this argument.
If we put your brain in a jar, would you be a person? Would you have normal consciousness? Honestly I don't know and I am not the person to ask.
But the brain is just a computer metaphor has flaws, particularly when you take account how poorly the brain is understood.
BENJY MOUSE Yes. To extract the question.
ARTHUR You can't have my brain. I'm using it.
FRANKY MOUSE Barely.
BENJY MOUSE We can replace it if you think it's important. An electronic brain maybe.
FRANKY MOUSE A simple one should suffice. Who would know the difference?
ARTHUR I would!
FRANKY MOUSE No, you wouldn't. We could program you not to.
The idea that we could have a silicon equivalent at all is already a huge assumption. For starters, the silicon equivalent cannot grow or kill neural connections or neurons as needed. Biological systems have a number of properties that silicon systems just don't at the moment.
Everything is a conspiracy here.
Of course it also reveals how unethical these actors are.
This may be an odd reference, but I remember this song and dance with Roomba Vs. Roborock. "Roborock stole our IP!!!" but the problem was that song was slightly believable in the very early days, then Roborock started releasing products far better than anything Roomba had, but the tune in the US never changed. Literally for years after Roborock's products were two generations better than Roomba, but somehow they "stole" that?
China has money, talent, and cheap electricity/superior infrastructure. The US's big AI built their products on stolen copyright material and social media scraping. China likely got a catchup boost via distilling, but recently a lot of their AI ventures have come fully online, and they've just throwing tons of compute/power at the problem, and we're starting to see legitimately competitive products without shortcuts/cheating.
But the US won't change its tune; and that's convenient because they don't really want to compete, but rather roll out protectionism and fear per normal.
Z.ai found a way to crack/overcome the cryptography that safeguarded Claude/Fable's thinking tokens in Claude Code, so Z.ai had the ability to peak into decrypted thinking tokens that Fable produced in Claude Code. Z.ai also shared their findings with other big LLM companies like Moonshot AI, which had 20,000+ accounts speed running through Fable to get their hands on as many high quality thinking tokens as possible for distillation.
https://github.com/rgreenblatt/which_claude_is_k3/blob/main/...
https://x.com/alex_prompter/status/2079625670918046048
Like all rumors, who knows what's true or what's not.
edit: the Anthropic employee downvotes are coming in now. This was +8, now it's +3.
We’re already competing on distills—they exist. The question is whether America will actually compete or if it will lean into its populist tendency and cede the market globally to China.
The sad fact here is that OpenAI and Anthropic are only looking for a max 1-year runway (Anthropic's is only a few months away) so that they can IPO and their VC investors cash out immensely. They do not care about AI safety or fairness, only to appease their VC overlords. But the public is being lied to and pitched this narrative that "open weight AI models are all bad".
This comment will be flagged soon and disappear just like anything else that does not support their "Trumpish" view of what the world should be.
Reagan and the Chicago school intentionally shifted antitrust and IP law into monopoly mode (Bork and Bowman+Posner). The Consumer Welfare Standard is an obvious fig-leaf over Robber Baron talking points as defeated earlier by Brandeis, the Nine No-Nos (lol) indicate that the failure modes of modern IP policy weren't just understood but contained, and then containment was turned off. Trade is a messier story but Triffin articulated the distributional problems in the 60s so nobody gets to claim ignorance.
The bad news is that there is more corruption that I thought, but the good news is that the problems are more solvable than I thought.
That, and usage was also half.
Given the fact that we seen not any evidence beyond people "see, it one shot X game looks similar to Fable" (a lot of one shots look similar to other models).
You expect to see something from Fable its restrictions/fallback behavior popping up in K3. You do not filter out all that behavior. Especially on such a short notice.
For me the milk example, where somebody asked if some EU cheese's had type A/B milk type (they are allergic), triggering bio restriction prompts. That just shows you how hard it is to find all those restrictions because its not in obvious spots where Fable goes off.
Yet, not a single person has shown this behavior in K3...
The funny thing is, other Chinese companies can train on K3 when the open weight are released next week.
So even if K3 was or was not trained on Fable, they already have a alternative to get their models to the same level.
1. "Eureka moments" often pop up almost out of nowhere, this suggests that there's processing going on at a non-verbal, non-token level.
2. The fact that we think in language is an evolutionarily new phenomenon. Plenty of "smart" animals are capable of complex behaviors and learning without any kind of language "tokens". There's no reason to think that human reasoning is different and somehow confined to "tokens" instead of being a broader system. Consider language itself. The human species created it somehow out of nothing. We "tokenized" our cognition to be able to communicate. But I would argue this was driven by our social needs, not our cognitive needs and the human ability to think does not require tokens at all.
Edit:
Finally, consider how limited language really is. Even for something simple, like pain, we don't even have a way to communicate objective reality to another person. We can say "it hurts" and use metaphors: "stabbing pain", "dull pain", etc. But the experience of pain can't really be communicated to another person. It's even worse once you remove non-verbal cues like facial expressions and body language - pain just becomes a vague concept on the page.
Meanwhile, our brain is perfectly capable of perceiving the slightest changes in pain and altering our behavior accordingly. Now extend this to other human experience.
I think the best analogy here is that treating human cognition as a problem of processing "tokens" is confusing the map for the territory.
Expect the tap on open models to slow and eventually disappear.
They don’t exist for altruistic reasons. It’s peculiar to current conditions.
Open weights models are already great, and the ones that are already released will continue to be just as good as they are today even if newer versions are released at a different speed in the future.
If your principles are only available on display when there is no consequence, they arnt your principles, they are a mask.
If you want AGI to arrive sooner, distills force you to move your compute off the frontier and will necessarily delay it significantly.
It's a question of what you actually want out of the AI race.
Yeah, I don’t buy this is happening. I’m open to being wrong. But folks who believe that can put their money where their mouths are in terms for stumping up cash for the training.
> distills force you to move your compute off the frontier
Versus moving it to inference?
There were so many security issues to patch in recent months that some open source projects stopped accepting them. I'm not sure how people still think Anthropic was lying.
the only one that was publicly admitted to and which you happened to have read about.
even prior to ai, most companies and organizations would never admit to being breached unless irrefutable proof of the breach came to light. in other cases it is just immediately classified and the public finds out 20 years later (or, depending on the country, simply covered up).
"A model being tested broke out of its sandboxing and hacked a production system of a different organisation using a previously-unknown vulnerability, in order to steal test results, and that proves anthropic were worried for no reason months ago about future models being security issues" ...?
Anthropic have released a lot of security patches in recent months, so many that many open source maintainers are facing burnout (google it). It's public knowledge that they weren't lying, you can look at the code. Their ridiculous guardrails were so that projects could patch and prepare for more models like this one coming out.
But companies do not care about social good, art, and other things outside of financials. A company is a generalization of sorts of all the people within the company who have influence over it's direction. 100% of people working for companies have a profit motive. That's why people work for for companies.
These people are the antithesis of "America is a Shithole", and they know it, so they keep their head down and point at the top 1% like everyone else. It's virtuous and pure to do so (and of course keeps the heat off them, the wealthiest contingent). If you isolate this group and compare them globally, it spells out why for decades so many white collar workers wanted to be in the US rather than Europe or the East.
Almost certainly not in the long term. If AI is a productivity enhancer, which I think we have evidence for, then the American economy’s heft (and continuing dynamism) get a tailwind. The minority of the new, leveraged economy getting hit will be a short-term painful adjustment, but it need not be anything more.
It almost certainly will. AI dominates growth. It doesn’t dominate the actual economy. Most spending and investment is entirely decoupled from it.
Given it sounds like you've already done this, care to share the results you got? When I meet peers from the US, it still seems like things are more rough there, regardless of class, compared to in other countries, but this is of course based on my personal experience, would love to see concrete data about it.
It is hard to fire employee, especially when the company is doing good, yes.
It is a bit slow and administratively annoying to start a business in some countries, yes, though in those you can usually start operating while starting AND there are workarounds which is why there isn't enough pressure to stop it (eg, the german "pre made company shell" market is how you actually do it instead of making your own).
Of course, distillation is also fair use under Copyright law.
Copyright was never meant to be a moral framework. It was always a practical framework designed to incentivize publishing that would ultimately pass into the public domain. Everybody seems to want to attribute some sort of moral weight to it though; the idea that people are naturally entitled to certain rights over things they've published. That idea would be totally alien to the people who designed the Copyright system in the first place.
> [the United States Congress shall have power] To promote the Progress of Science and useful Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries.
The law may see this as no issue, as the law cares more about protecting power, but that does not mean it isnt immoral.
I am completely willing to accept that "this is fair use" for any company that publishes the LLM weights, i.e. the result of processing all the copyrighted work, because they have performed a public service with this.
But when the so-called "fair use" was a method to transform public data into private data that they guard and claim that any access to it would now be IP theft and which they use to obtain huge profits, that does not look like fair use to me.
The whole point of fair use (which courts have so far ruled AI training is) is that you don't have to ask for permission or pay them for it.
Is that settled law? I doubt it.
And IMHO, AI training violates the spirit of fair use. It's not really a method of criticism or commentary. It's a system to use people's own work to undermine their ability to economically subsist on that work.
Though how about this for a proposed exception: you can AI train on anything as fair use: only if release your model and weights public domain open source.
It seems to be the fairly consistent approach of trial courts addressing the question under different soecific fact patterns in different contexts; its not “settled law” in the sense of nationally binding precedent (which would take either a Supreme Court ruling kr separate appellate rulings in every circuit).
> And IMHO, AI training violates the spirit of fair use. It's not really a method of criticism or commentary.
Plenty of transformative uses that have been held to be fair use are not criticism or commentary, and AI training is a transformative use where, for any individual work used, the end product is both a very different class of work and the used work indiviudally has very small impact on the final work.
> It's a system to use people's own work to undermine their ability to economically subsist on that work.
Courts seem to disagree that this is generally the case with AI training as such. (And AI training being fair use would not make the use of models to create works that would otherwise be infringing copies with that function through inference any less infringing.)
> Though how about this for a proposed exception: you can AI train on anything as fair use: only if release your model and weights public domain open source.
You are, of course, free to try to convince Congress to amend copyright law to apply that rule (though since the current statutory form of the fair use rule is itself a legislative adoption that follows pre-existing court rulings on fair use as a Constitutional limit on the copyright power stemming from the First Amendment, Congress may not actually have the power to narrow it that way.)
That just seems like a cope unless you have actual evidence that the lower/appellate courts have misruled. And no, "I don't like the ruling because [all the reasons AI is bad]" doesn't count, you need actual legal justifications, preferably from legal experts. Not to mention that even if the supreme court ruled on it, it's not really "settled", eg. Roe. v. Wade and Humphrey's Executor v. United States being overturned
No, it means lower court judges get overruled all the time and it's not like the courts and law always functions as some dispassionate applicators of some fixed framework. It's not settled until the process gets worked much farther than it probably has.
No, it's not the same thing. Contrary to what many people think, "fair use" isn't something you can invoke to do whatever copyright infringement you want. The judge is supposed to consider several factors, one of which is whether the work was "transformative". In google's case it was offering search results. Internet archive was operating a "digital library" (aka. a filesharing site). Whatever you hate about AI companies sucking up electricity and displacing jobs, they're certainly more transformative (and arguably more transformative than even google search) than whatever the internet archive was doing.
If I run a book through AES, that’s pretty transformative too!
And has this been tested in court? After all, you see people uploading tv shows on youtube, then pasting a snippet of fair use in the description. That doesn't make it true. If anything, the unfavorable ruling for internet archive suggests libraries were incorrect with their interpretation of the law.
>2. Google provided access to the whole book, that’s why they got sued.
No it didn't. From wikipedia:
"For works still under copyright, Google scanned and entered the whole work into their searchable database, but only provided "snippet views" of the scanned pages in search results to users."
