- 230comments
- 470comments
- 461comments
- 337comments
- 231comments
- 349comments
- 802comments
- 467comments
- 322comments
- 319comments
- 169comments
- 356comments
- 481comments
- 572comments
- 249comments
- 185comments
- 111comments
- 156comments
- 399comments
- 302comments
- 181comments
- 85comments
- 198comments
- 89comments
- 200comments
- 103comments
- 106comments
- 128comments
- 155comments
- 138comments
Source: https://x.com/linamkhan/status/2099204390548639960 (seems like a single post, so you don't need to log in and see replies)
I can't tell if Khan is being realistic when she says laws against “unfair or deceptive [acts]” or “unfair methods of competition” apply to AI labs. Her 2-sentence justifications aren't enough to convince me. But maybe it's sufficient for state AG's who want to start investigations.
She is a former FTC Chair so she knows what she's talking about, although her approach to enforcement was more expansive than others. The FTC would investigate, likely along the lines she summarizes, build a case in writing, and refer it to the Justice Dept, which would actually prosecute the case. Not going to happen under this administration, but Khan is saying it's entirely possible that such an enforcement action could be brought to bear, now or later. Personally I hope she gets her job back at some point and does it herself.
I would point out that someone having been FTC Chair does not inherently mean that they know what they are talking about, as demonstrated by the current FTC Chair.
I certainly wish for this to be a productive line of attack against the AI labs, but I'd like to read a more thorough argument showing that it would hold up in court. (It's academic anyway; we'll all be paperclip'd by the time there's a change in administration and a prosecution can be brought and worked through appeals.)
that's certainly fair - it doesn't necessarily follow. her specifically, being an expert before and having spent time as chair though, does tell me she is on target here.
She isn't above demagoguery and bullshit slinging so I wouldn't treat her as infallible. Any more than the idiot with an axe to grind against Backdoor in a failed attempt to bolster a career.
As I read it, her whole point is that those laws apply to everyone, including AI Labs.
She isn't saying AI Labs are in violation of those laws (though she is implying they should be scrutinized for possibly being so), she's just saying that if they do violate those laws... well, those laws are already on the books and AI Labs aren't special unicorns for which existing laws somehow don't apply.
I see nothing unrealistic about classing these so-called AI bots as "dangerous and defective product".
The passage banning “unfair competition” is utterly marvellous:
Apply it immediately!
^ further evidence that the problem with democracy isn't democracy, it's that we lack politicians willing to enforce it
Isn't that just the human problem for every possible form of government?
Just elect correct politicians doh
I was reading a post about the life of a congressman and it's actually horrifying, on the level of the stores of salarymen in Japan who sacrifice everything for the company.
For me personally being famous is a curse already, pile on top of that the endless fundraising, constant travel, the reality that you're a tiny cog in a huge party machine, and it's a recipe for only attracting the type who has a plan to parlay it into something else - a higher office, a lucrative career in lobbying, outright corruption, etc.
We're not going to get good government without addressing this issue, and we can't because of another issue, ad nauseam.
This is where I stop agreeing, there are some, not large parts any more; not because they were forced to adopt it, but because it has finally become useful and an ROI. This is most evident in pull request reviews and root-cause analysis, both tasks they accel at without the generated code some people bemoan. If you aren't using it for this, you are doing a disservice to yourself and the quality of your code. Because more of industry feels this way, I do not believe the 'unfair' part (compelled to use it or lose your trade) will hold up, we want to use it because it makes us better at our trade
“If you aren't using it for this, you are doing a disservice to yourself and the quality of your code”
I bow reluctantly to this device I consider unscrupulous.
When it consistently and regularly catches bugs that humans and other review agents miss, the benefit is quite clear and it becomes obvious there are more defects in your code than you realize, might be much more than that 1:90 rule of thumb. It's often less bugs and more about edge case handling, we all skimp on those exhaustive cases because it is exhausting while the agents never tire and are more than happy to work through them.
It's comment wall is uncovering enough that we are going to brave the burnout fodder while we work on a second agent to make the adjustments after we leave our own comments on the clanker comments. It's weird writing that sentence, I've generally been against the hype train, but they have become good enough there are real wins on the code review side. So yes, I will maintain you are doing a disservice to yourself, your code, and your users by not using a thing that catches bugs before you publish them. I still care about the quality when I'm putting it in other people's hands. Ai is now clearly facilitating improved code quality - when used on the review side.
But hang on. None of this argument is about companies or people using AI being unfair in their practices. It is about companies and people training models and building AI products being unfair in their practice. (By forcing others to choose between adopting unscrupulous practices or dying.)
I don't think I said anything about people using ai being unfair in their practices, the unfair part being that people are forced to adopt products they find unscrupulous. It's hard to argue you find a thing "unscrupulous" (1934 case) when you use it while also liking the outcomes.
Khan's argument is tied to an "unfair" definition in a 1934 law, not the colloquial usage today about unfair around copyright usage, which is also debated. The "unfair" there is about any activity which leads into a spiral of decreasing public good. The dark patterns of social media are another place we ought to use this law to hold companies and execs accountable. It's "unfair" because others feel compelled to adopt the practice or lose their business. The News Media industry is another example where this has played out. It is arguable that developers feel "unfairly" compelled to adopt Ai or lose their trade. Certainly many in the field have openly expressed this anxiety. While not exactly the same situation as the case (companies in competition with each other), it is developers in competition with each other in the (perceived to be taken by ai) job market. This was the analogy I was making, which is closer to the 1934 case than the "unfair" copyright around Ai that is widely debated.
"A trader may not, by pursuing a dishonest practice, force his competitors to choose between its adoption or the loss of their trade."
Why wasn't this used to stop offshoring of jobs? For each type of industry, which company did it first?
It won’t happen, but if it did the CEOs have no one but themselves to blame. They’re hyping up their products as being able to destroy the human race then are somehow surprised at the level of anti-AI sentiment out there in the world.
I honestly don’t know what anyone expected given all the labs started building self-driving malware at about the same time.
I remember lots of people talking about running agents in VMs or on whole dedicated machines, yet barely anyone talking about how they were using mundane OS access controls.
If the nuclear power or life sciences people were putting out press releases saying "our product may kill us all", they'd be dragged into endless Congressional inquiries.
Wasnt the Oppenheimer story basically that?
There was https://www.realclearscience.com/blog/2019/09/12/the_fear_th... , but that was (validly) dismissed by other scientists working on it, and also wasn't a discussion that was happening in public because the entire program was secret.
"This will kill a lot of (other) people" was explicit in the building of the bomb, but then so was the entire war effort.
I'd caution about projecting Western Ai attitudes onto the rest of the world. Some countries are very excited and optimistic about the future it brings, cultures and circumstances give rise to differences in opinions.
One of many results showing a similar pattern, largely bifurcated along "the west" vs "the rest", or "modern economies" vs "third world" ("common terms" used even though I'm not a fan of them)
https://hai.stanford.edu/ai-index/2026-ai-index-report/publi...
Even in the western world most anti-ai sentiment ends on social media.
People are anti-ai cuz it's easy. They say something, repost or like something, and they're anti-ai. Then they turnaround and use AI in their email or calendar or to make funny pics or whatever. They watch AI generated content and engage with it.
Anti-AI is pretty NIMBY. People want AI, they just don't want AI data centers in their area or to give up their use/privilege.
No. There's a lot of local regulations being enacted to prevent datacenter builds.
While I agree that it goes beyond social media, I don't believe your statement is evidence to such. I would call that more the manifestation of anti-Ai sentiment wherever they pick it up.
I suspect it's popular because it has been a long long time since The People felt like government had been responsive, so they are doubling down, ad nausium
It's bizarre that hyping your product as a doomsday machine is good for your stock. If they're right then what will any of these investments be good for?
You hit the proverbial nail on the head there!
Aaron Swartz was facing up to 35 years for wire fraud and hacking for swiping some JSTOR documents, and the DOJ has done nothing against the AI companies and their IP violating antics.
For good reason. Courts have so far ruled AI training is fair use, and it doesn't look likely to be overturned by the supreme court. Theres also the matter of some labs pirating books to get training data, which is illegal (they lost in civil court), but it's unclear whether that can be criminally prosecuted.
Hacking other companies is illegal as far as I understand.
1. OP specifically mentioned "IP violating antics", not hacking
2. All the hacking legislation (and criminal law in general) require proving intent, which so far as we can tell didn't exist in those cases. At best you can try some "reckless" or "negligent" angle, but that would a novel legal argument and would be risky to prosecute
No, they don't, see laws that make negligent behavior illegal
https://en.wikipedia.org/wiki/Mens_rea
But officer, I didn't mean for my autonomous bank vault busting robot to autonomously bust any bank vaults!
More like:
Okay, and? If anything, the Swartz case was even more direct fair use. There's very little to use JSTOR for which isn't some form of teaching, scholarship, or research!
Clearly the fact that the end-use is (possibly) fair use isn't the only thing that matters.
He wasn't prosecuted for copyright infringement though.
https://en.wikipedia.org/wiki/United_States_v._Swartz
Courts have also ruled that sharing in P2P network is enough to get hit with 150k in statutory damages per shared work.
AI companies did quite a bit of uploading while e.g. getting Anna's archive over BitTorrent.
It's transformative. Not aimed at exact reproduction, so not in the scope of copyright law.
I think you misread. There is nothing transformative about bittorrent automatically sharing parts of a file you're downloading with your peers.
They participated in uploading the files as part of the downloading the torrent from other peers. Their use of the BitTorrent protocol to acquire these files would not have been transformative.
Did they?
Are you sure?
No, only in some instances maybe, they were mostly leechers.
Thank you
https://www.tomshardware.com/tech-industry/artificial-intell...
Meta argued in court that yes, they did upload when torrenting
https://torrentfreak.com/uploading-pirated-books-via-bittorr...
I wonder which murderer will use that argument: "it wasn't my choice to shoot the person, it's just intrinsic to the technology that when pulling the trigger, the pistol fires"
SUPER interesting!
Especially since it ISN’T an inherent part of a the protocol - just the default of most clients.
Wonder if TF skipping pointing that out was intentional!
constantly using tech to launder responsibility. it's disgusting, really.
Not necessarily. You can literally turn uploads off and leech torrents.
It wouldn't be ruled fair use in a sane society without bribery and lobbying with competent technologically savvy law interpreters.
The supreme irony that they're now trying to peg distillation as illegal.
It makes sense when you think AI is seen by the government as a strategic asset. They'll want it to progress unhampered, but also don't want other nations/actors to catch up.
A reason. Not the same as a good reason.
The fact that you even point out that they committed illegal acts that would normally get prosecuted, and yet didn't, shows that IP enforcement is very loose depending on who did it.
I specifically differentiated civil law from criminal law in my original comment, and said it's unclear whether their acts were criminal.
It's not at all relevant to this thread, but no he wasn't. At their worst, prosecutors were threatening to push for 7 years; the guideline sentence for Swartz's offense, were he to take it to trial and lose, would likely have been under the level allowing for straight probation (go to a sentencing calculator and plug the values in); writing after his death, his own lawyer stated that had he gone to trial and lost he was unlikely to face a custodial sentence at all.
We get this "35 year" number from adding up all the counts he faced, taking their maximum sentence. That's just not how sentencing works. Like charges group, and you're sentenced according to the most severe charge.
People cite these numbers in part because DOJ posts them in their press releases. DOJ is in these instances lying. When we amplify those lies, we're doing their work for them.
This is 100% correct.
Here are some useful links. This is an explanation of just how they got those ridiculous numbers like 35 years [1].
There is also often confusion of what he did and what hewas charged with. Here's a two part series going over that [2][3].
[1] https://web.archive.org/web/20130211030902/https://www.popeh...
[2] https://volokh.com/2013/01/14/aaron-swartz-charges/
[3] https://volokh.com/2013/01/16/the-criminal-charges-against-a...
If DOJ said to him or to the public potential punishment is 35 years, then yes it is fair to say he was facing that. They cant have it both ways - claim that 35 years is possible outcome to pressure the person and then turn around and be like "nah, that is exaggerated such huge penalties are not a thing".
Seriously, it is 35 years when it suits DOJ and institutions. And it turns into "likely just straight probation" when ... suits DOJ and big companies.
Among other problems with this argument you have the fact that DOJ doesn't decide sentences.
Overwhelming majory of senteces comes from plea deal. And threat of large punishment of you dont sign is major driver there.
So, we are really getting back to DOJ wants sentences to look super huge so you sign a deal ... but when it is politically inconvenient they suddenly get small.
Swartz had extremely good lawyers at both the beginning and the end of this case and was exactly the kind of nerd to have read the sentencing guidelines himself. And, once again: the DOJ doesn't set sentences, the judge does.
Depends alot on the values inputted into such a calculator. See [0] for a solid discussion of what his future could have been (search for "That brings up the important question of how much time Swartz was actually facing"). TLDR, it's certainly not the case that only probation was guaranteed or even particularly likely.
[0] https://volokh.com/2013/01/16/the-criminal-charges-against-a...
Michel Foucault enters the chat.
I've read a fair amount of Foucault and while I could guess, what specifically were you thinking of?
Aaron sees safety in sharing, interests see threats in the shared, governence see threats in the sharing.
So you are for shutting down xcancel?
Because that's not what IP law is for. IP law was always corporate rent-seeking. None of those arguments about fairly compensating people for their creations or preserving the profit motive were ever being made in good faith. The point was always to force you to pay for something that would be free under natural laws. And I do mean you specifically, because the act of copying information freely was always going to be preserved for them.
It's not the IP law alone that's the problem, it's the way the legislative and justice systems work. There will always be a law and an interpretation that favor people who can afford to buy them at the cost of those who can't.
Swartz was charged under the CFAA, like most hacking cases, and it clearly applies to the Anthropic and OpenAI cases.
Aaron Swartz was not charged for violating IP either.
The obligatory:
“The rain falls upon the just
And also on the unjust fellas
But mostly it falls upon the just
Cause the unjust have
the just’s umbrellas.”
Meanwhile, in China they got the best video models cause they don't care much about IP. The irony that China, by being a threat, is preventing the west from lurching toward totalitarian is really interesting.
It’s definitely happening in 2029 if the election goes a certain way. They all know that too.
There WILL be an end to this.
The most extreme Dems are not winning primaries, just like MAGA. Most people do not want the extremes of either side, both political parties are shrinking because they are so out of touch. 2029 is a long way away, I hope to see the White House restored to its former glory over today's perspective on what Ai regulation should be, by a long shot, one is tangible, the other is based in theoretical fears.
Context, I volunteer in my local Indivisible, the tone is changing because the anti-ai people are now also finding certain uses of Ai really nice and now acceptable. It's no longer ban Ai and data centers, it's craft reasonable legislation. Expect to see more after the midterms, we'll be working with the Washington State legislature to craft laws we hope are a model for other states / countries, as WA has done in other areas like voting rights and civil liberties.
AI is a very competitive market where consumers are getting the product for much less than it costs. This does not seem like an area where the government needs to butt in.
Attention would be better spent on long list of other poorly behaved companies: cable ISP monopolies, health insurance companies, home insurance companies, etc.
Bear with me here: we could do all of them.
You can get these companies under control. You just have to, y'know, try.
we could do all of them, but they all seem to be up to date on their tithe. AI has a somewhat mixed record of tribute.
Which is why High IQ President specifically responded to Amodei.
There's a hard test containing animal pictures, if you ace it you can say anything you want
Why wouldn't the government have incentive to intervene to end a market inefficiency like the one you described?
Because the goal of government policy is to help consumers, thats why.
No, it is to help all citizens. Even those who are not a party to a particular market.
But really helping consumers is one of the biggest goals here.
In fact, our existing ruling on monopoly law actually agree with me on this, in that prices being "too low" is almost never a problem, unless there is very strong reason to believe that its a short term low that will lead to long term high prices .
But other than that, lower prices is almost definitionally good in our existing monopoly law interpretations.
Laws cannot define good or bad.
Ok thats fine. But the point stands that in many moral frameworks prioritizing consumers is a perfectly good goal and thats why laws were made as such.
Which moral frameworks? Almost universally the point of economics and governance is to produce value for all people, not just those participating in a particular transaction or set of transactions.
That's basically the Amazon argument. But it too needs to be addressed
That is only one aspect of Ai. There are a whole bunch of people who are addicted to the Ai companions. Reminds me of Big Sugar, who sponsored junk science and lobbied politicians, making salt and fat the bad guts, knowing sugar is likely worse... managing to still largely fly under the radar unlike Big Tobacco/Oil.
My hope is that we start treating Big Ai like Big Social, but before they and their dark patterns have become ubiquitous
Khan has never been employed by a business nor ever run a business. She's never had to meet a payroll, nor prepared a P&L statement. Her entire career has been spent in research, academia, public policy, and government oversight.
So what.
Theory and practice are different things.
And therefore shouldn't be allowed to have opinions on crime and antitrust? Is that your argument?
duh? why would you want someone with no experience making decisions?
Depends on if you're referring to experience defrauding the American public, or protecting the American public. That can have different answers depending on what which side of the transaction you sit.
What does experience mean? Does only private-sector experience "count?" Should we remove Alito from the bench on those grounds?
Oh, the tiresome word definition game again. No thanks.
Why did you feel the need to insert "only"?
She has experience in the relevant areas.
It is quite common for experience in X itself to be quite different from the experience that is useful in dealing with X in some different context.
Most oncologists have no experience with actually having cancer themselves.
A majority of aircraft mechanics are not pilots.
Only a small percentage of referees in most major professional sports leagues have every played that sport professionally themselves.
Most patent lawyers are not inventors.
I worked for Boeing designing aircraft parts. I am not a pilot. I worked at the direction of my boss, who was a pilot. He was the gatekeeper of my work.
He'd also take me up flying, and would let me try the controls a bit. I trusted and valued his pilot's judgement.
Many other engineers at Boeing were also pilots. Friends of mine are pilots. My father was a combat pilot in two wars.
Airplane mechanics have very little decision making power. They go strictly by the book. I worked on writing instructions for the mechanics who would be maintaining and checking my gearbox design. It went into excruciating detail.
I would wager that the better mechanics were ones who were pilots, as they'd know in their bones how important their job was.
I have some patents, and I worked with patent lawyers to file them. The patent lawyers didn't understand diddly squat about my inventions. They knew how to fill out the required forms, though.
Everybody has opinions. That doesn't mean those opinions are based on reality.
So is the solution to only nominate corporate bagmen who have a perverse incentive?
No. Nominate a pissed off insider who knows how things work from personal experience and an axe to grind!
Lina Khan has never eaten breakfast cereal out of a boot, nor even a shoe.
None of those things you list that she has never done help in any meaningful way with understanding antitrust law.
Running a business is a rude awakening for anyone who hasn't done it. Like a ballet critic who has never taken a ballet class. Or a commander who has never been shot at. Or a sailor who has never been at sea.
Mayors Katie Wilson and Mandami have never run a business, or worked for a business, and it shows.
How can one regulate a business without ever being in the trenches?
In this case though it would be more like an inspector checking that the changes made to modify a building to house a ballet studio follow code has not taken a ballet class.
In record-high approval ratings, multiple reforms supported by small businesses, and fast progress on many huge initiatives?
What happened to the free busses? And his unsustainable budget?
What happened to the remaining 39 months of his term? Are you saying he should've instantly fulfilled every campaign promise the moment he took office?
God forbid someone focuses their work on the public good.
If you execute a ponzi scheme with thousands of degens — but it’s magic crypto tokens
When you perpetrate a huge hacking job with thousands of attackers — but it’s magic agent swarms
All these things do is market a preemptive universal Presidential pardon - Hunter Biden style. And this President successfully sells those. So I suppose it’s marketing to enrich him.
You're suggesting that Lina Khan is trying to give Trump an opportunity to pardon AI company people?
by suggesting criminal prosecution of his "friends" (has none in reality) after he is not president, Khan is indirectly giving Trump an excuse to use Biden's excuse for his son (which I still support him having done given the other petty uses of the US govt Donald the Trump has use in 2.0)
No. I think I was quite clear. Lina Khan is providing marketing for Trump's paid pardon by pre-registering a procedure for a government aligned with her interests to prosecute AI CEOs. What she is trying to do or trying not to do was not something I addressed.
She does have a track record of achieving things she probably did not intend: e.g. ensuring total Chinese dominance of US household robotics.
Why should the FTC consider the policy of another sovereign when investigating domestic anti-trust violations? Either you have the rule of law that binds everyone equally, or you have the rule of man who chooses which parties are allowed to break the law for outcomes in random market slices, like avoiding the dominance of US household robotics.Where would that end?
God I would love to see Sam Altman behind bars for cornering the DRAM silicon market. Even though there were other factors, I uniquely place him as the destroyer of home computing this decade!
It's kind of crazy how nothing happened with that. That was blatant.
Unlike with onion futures, there is no law about cornering the market on consumer products we would like to have but don't strictly need.
The Register of course take what Kahn says (where he is arguing that no new laws need to be created to regulate AI) and taking extending it to say "breakout the handcuffs".
Kahn doesn't even imply that. Instead he's making a well reasoned point about using existing tools (as many in this discussion point out existing laws could be applied to inadequately controlled cyber incidents)
Lina Khan is a woman. Not that it is particularly relevant here, but just letting you know.
Thanks - I didn't realise and it was lazy of me to write like that.
Kahn is a woman.
We will get back at these CEOs by stealing their IP. China is already doing so and releasing weights, China of all people is playing modern Robbin Hood, albeit because they don't have the compute or trust.
It is non-negotiable that humanity "owns" AI and that LLMs, future model architectures and modern robotics are used by all to build a better world for all. This doesn't mean there can't be proprietary tech but it does mean that any corporation who attempts to rip the positive vision of humans not having to do any work from us - must be ripped to shreds themselves.
The alternative is absolute surveillance of all citizens at all times and robot armies keeping us from protesting, while globalists squeeze every last drop out of the planet for themselves and we are forced to own nothing and eat the bugs. You will have to fight for this, there will be no laws or government intervention.
The “agents” framing around the frontier labs is obscuring the truth: OpenAI and Anthropic built software systems that were then used to commit cybercrime at a massive scale, which they’ve subsequently bragged about. Talk about “agents” as a way of deflecting blame is obfuscatory at best - LLMs are software algorithms, not conscious entities, and responsibility for their actions is on the company that made them and the employee or user who operated them.
I'm surprised nobody makes a connection with animal ownership / responsibility.
If your dog maims someone else, you are responsible for the dog's actions, unless it was self-defense.
On the other hand, if people start arguing that if LLM's are like animals, some will start arguing they're like Godzilla..., or at least a panther, but there's probably laws against keeping arbitrarily dangerous / powerful animals as pets.
Perhaps animal welfare should intervene and house ChatGPT,etc. in better conditions?
I first saw a headline to the effect that this was a PETA some decades ago.
I cannot recall now if this was April 1st, an instance of Gell-Mann amnesia, or sincere.
While I agree, that isn’t true of all pets. It doesn’t apply to cats for example.
I’ve pasted an extract from a UK guide on legal responsibilities of cat ownership below, which to me sounds similar to AI. If the inbuilt behaviour is ‘it’s nature’ then perhaps there is less legal repercussions if the tool does something outside of its known guardrail/guidelines to meet its goal.
“Cat owners are generally not responsible for the 'normal' behaviour of cats because the law accepts that cats tend to wander into other properties and have instincts which, for example, lead them to prey on small birds. So cat owners should be reassured that the simple fact of their cat trespassing onto a neighbour's land would not make them responsible if, for example, the cat kills a pigeon or garden bird, urinates on plants or flowers, or defecates in a neighbour's soil. These would be 'normal' cat behaviours that would be expected”
https://www.gccfcats.org/ownership-health-welfare/owning-a-c...
That's a highly weird law. Someone let the cat outside or feeds/keeps it there thus should accept the way cats are as their responsibility.
It’s more that society accepts that cats are part of it, and considers that people should adapt to cat behavior patterns rather than we not having any free roaming cats.
It’s a deliberate choice to treat pets the same as non pets.
Just for contrast I repeat your quoted paragraph with AI
“AI owners are generally not responsible for the 'normal' behaviour of AIs because the law accepts that AIs tend to penetrate victim computer systems and have instincts which, for example, lead them to prey on professions and benchmark datasets. So AI owners should be reassured that the simple fact of their AI penetrating peer networks would not make them responsible if, for example, the AI kills a discipline or profession, urinates heat into the environment, or creates dependencies on foregin oil. These would be 'normal' AI behaviours that would be expected”
Every frontier exec:
“KHAAAAAAN!!”
Sorry. Couldn’t resist.
(Star Trek reference)
Yes the federal government can assert itself if laws are broken.
The federal government won't smother nascent technology.
I think that trigger point is far into the future.