The criticism from large AI companies to this bill sounds a lot like the pushbacks from auto manufacturers from adding safety features like seatbelts, airbags, and crumple zones. Just because someone else used a model for nefarious purposes doesn’t absolve the model creator from their responsibility to minimize that potential. We already do this for a lot of other industries like cars, guns, and tobacco - minimize the potential of harm despite individual actions causing the harm and not the company directly.
I have been following Andrew Ng for a long time and I admire his technical expertise. But his political philosophy around ML and AI has always focused on self regulation, which we have seen fail in countless industries.
The bill specifically mentions that creators of open source models that have been altered and fine tuned will not be held liable for damages from the altered models. It also only applies to models that cost more than $100M to train. So if you have that much money for training models, it’s very reasonable to expect that you spend some portion of it to ensure that the models do not cause very large damages to society.
So companies hosting their own models, like openAI and Anthropic, should definitely be responsible for adding safety guardrails around the use of their models for nefarious purposes - at least those causing loss of life. The bill mentions that it would only apply to very large damages (such as, exceeding $500M), so one person finding out a loophole isn’t going to trigger the bill. But if the companies fail to close these loopholes despite millions of people (or a few people millions of times) exploiting them, then that’s definitely on the company.
As a developer of AI models and applications, I support the bill and I’m glad to see lawmakers willing to get ahead of technology instead of waiting for something bad to happen and then trying to catch up like for social media.
self regulate? big tech company? pfft right we all know how that goes
the people who are already being victimized by ai and are likely to continue to be victimized by it are underage girls and young women.
The bill, passed by the state’s Senate last month and set for a vote from its general assembly in August, requires AI groups in California to guarantee to a newly created state body that they will not develop models with “a hazardous capability,” such as creating biological or nuclear weapons or aiding cyber security attacks.
I’ll get right back to my AI-powered nuclear weapons program after I finish adding glue to my AI-developed pizza sauce.
this is where the AI hype has taken us
Cake and eat it too. We hear from the industry itself how wary we should be but we shouldn’t act on it - except to invest of course.
The industry itself hyped its dangers. If it was to drum up business, well, suck it.
The idea of holding developers of open source models responsible for the activities of forks is a terrible precedent
The bill excludes holding responsible creators of open source models for damages from forked models that have been significantly altered.
Small problem though: researchers have already found ways to circumvent LLM off-limit queries. I am not sure how you can prevent someone from asking the “wrong” question. It makes more sense for security practices to be hardened and made more robust
I had a short look at the text of the bill. It’s not as immediately worrying as I feared, but still pretty bad.
https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202320240SB1047
Here’s the thing: How would you react, if this bill required all texts that could help someone “hack” to be removed from libraries? Outrageous, right? What if we only removed cybersecurity texts from libraries if they were written with the help of AI? Does it now become ok?
What if the bill “just” sought to prevent such texts from being written? Still outrageous? Well, that is what this bill is trying to do.
Not everything is a slippery slope. In this case the scenario where learning about cybersecurity is even slightly hinderedby this law doesn’t sound particularly convincing in your comment.
The bill is supposed to prevent speech. It is the intended effect. I’m not saying it’s a slippery slope.
I chose to focus on cybersecurity, because that is where it is obviously bad. In other areas, you can reasonably argue that some things should be classified for “national security”. If you prevent open discussion of security problems, you just make everything worse.
Yeah, a bunch of speech is restricted. Restricting speech isn’t in itself bad, it’s generally only a problem when it’s used to oppress political opposition. But copyrights, hate speech, death threats, doxxing, personal data, defense related confidentiality… Those are all kinds of speech that are strictly regulated when they’re not outright banned, for the express purpose of guaranteeing safety, and it’s generally accepted.
In this case it’s not even restricting the content of speech. Only a very special kind of medium that consists in generating speech through an unreliably understood method of rock carving is restricted, and only when applied to what is argued as a sensitive subject. The content of the speech isn’t even in question. You can’t carve a cyber security text in the flesh of an unwilling human either, or even paint it on someone’s property, but you can just generate exactly the same speech with a pen and paper and it’s a-okay.
If your point isn’t that the unrelated scenarios in your original comment are somehow the next step, I still don’t see how that’s bad.
Restricting speech isn’t in itself bad,
That’s certainly not the default opinion. Why do you think freedom of expression is a thing?
Oh yeah? And which restriction of free speech illustrating my previous comment would is even remotely controversial, do you think?
I’ve actually stated explicitly before why I believe it is a thing: to protect political dissent from being criminalized. Why do you think it is a thing?
And which restriction of free speech illustrating my previous comment would is even remotely controversial, do you think?
All of these regularly cause controversy.
I’ve actually stated explicitly before why I believe it is a thing: to protect political dissent from being criminalized. Why do you think it is a thing?
That’s not quite what I meant. Take the US 2nd amendment; the right to bear arms. It is fairly unique. But freedom of expression is ubiquitous as a guaranteed right (on paper, obviously). Why are ideas from the 1st amendment ubiquitous 200 years later, but not from the 2nd?
My answer is, because you cannot have a prosperous, powerful nation without freedom of information. For one, you can’t have high-tech without an educated citizenry sharing knowledge. I don’t know of any country that considers freedom of expression limited to political speech. It’s one of the more popular types of speech to cause persecution. Even in the more liberal countries, calls to overthrow the government or secede tends to be frowned on.
Do they really? Carving into people’s flesh causes controversy? The US sure is wild.
Even if some of my examples do cause controversy in the US sometimes (I do realize you lot tend to fantasize free speech as an absolute rather than a freedom that - although very important - is always weighed against all the other very important rights like security and body autonomy) they do stand as examples of limits to free speech that are generally accepted by the large majority. Enough that those controversies don’t generally end up in blanket decriminalization of mutilation and vandalism. So I still refute that my stance is not “the default opinion”. It may be rarely formulated this way, but I posit that the absolutism you defend is, in actuality, the rarer opinion of the two.
The example of restriction of free speech your initial comment develops upon is a fringe consequence of the law in question and doesn’t even restrict the information from circulating, only the tools you can use to write it. My point is that this is not at all uncommon in law, even in american law, and that it does not, in fact, prevent information from circulating.
The fact that you fail to describe why circulation of information is important for a healthy society makes your answer really vague. The single example you give doesn’t help : if scientific and tech-related information were free to circulate scientists wouldn’t use sci-hub. And if it were the main idea, universities would be free in the US (the country that values free speech the most) rather than in European countries that have a much more relative viewpoint on it. The well known “everything is political” is the reason why you don’t restrict free speech to explicitly political statements. How would you draw the line by law? It’s easier and more efficient to make the right general, and then create exceptions on a case-by-case basis (confidential information, hate speech, calls for violence, threats of murder…)
Should confidential information be allowed to circulate to Putin from your ex-President then?
Seems a reasonable request. You are creating a tool with the potential to be used as a weapon, you must be able to guarantee it won’t be used as such. Power is nothing without control.
I am pretty sure no one has ever built a computer that can’t be shut off. Somehow someway.
How is that reasonable? Almost anything could be potentially used as a weapon, or to aid in crime.
This is for models that cost 100 million dollars to train. Not all things are the same and most things that can do serious damage to big chunks of population are regulated. Cars are regulated, firearms are regulated, access to drugs is regulated. Even internet access is super controlled. I don’t see how you can say AI should not be regulated.
AI is already regulated. Just because something is new (to the public) does not mean that laws don’t apply to it. We don’t need regulation for the sake of regulation.
There’s a lot of AI regulation that may become necessary one day. For example, maybe we should have a right to an AI assistant, like there is a right to legal counsel today. Should we be thinking about the minimum compute to assign to public defense AIs?
This is for models that cost 100 million dollars to train.
Or take a certain amount of compute. Right now, this covers no models. Between progress and inflation, it will eventually cover all models. At some point between no and then, the makers of such laws will be cursed as AI illiterate fools, like we curse computer illiterate boomers today.
Think about this example you gave: Cars are regulated
We regulate cars, and implicitly the software in it. We do not regulate software in the abstract. We don’t monitor mechanics or engineers. People are encouraged to learn and to become educated.
Of course you regulate software in the abstract. Have you ever heard of the regulations concerning onboard navigation software in planes? It’s really strict, and mechanics and engineers that work on that are monitored.
Better exemple: do you think people who work on the targeting algorithms in missiles are allowed to chat about the specifics of their algorithms with chat gpt? Because they aren’t.
I guess let’s deregulate guns then. Oh wait.
This bill targets AI systems that are like the ChatGPT series. These AIs produce text, images, audio, video, etc… IOW they are dangerous in the same way that a library is dangerous. A library may contain instructions on making bombs, nerve gas, and so on. In the future, there will likely be AIs that can also give such instructions.
Controlling information or access to education isn’t exactly a good guy move. It’s not compatible with a free or industrialized country. Maybe some things need to be secret for national security, but that’s not really what this bill is about.
Everyone remember this the next time a gun store or manufacturer gets shielded from a class action led by shooting victims and their parents.
Remember that a fucking autocorrect program needed to be regulated so it couldn’t spit out instructions for a bomb, that probably wouldn’t work, and yet a company selling well more firepower than anyone would ever need for hunting or home defense was not at fault.
I agree, LLMs should not be telling angry teenagers and insane righrwungers how to blow up a building. That is a bad thing and should be avoided. What I am pointing out is the very real situation we are in right now a much more deadly threat exists. And that the various levels of government have bent over backwards to protect the people enabling it to be untouchable.
If you can allow a LLM company to be sued for serving up public information you should definitely be able to sue a corporation that built a gun whose only legit purpose is commiting a war crime level attack with.
that is not the safety concern.
Guns aren’t a safety concern. Ok then
The safety concern is for renegade super intelligent AI, not an AI that can recite bomb recipes scraped from the internet.
Damn if only we had some way to you know turn off electricity to a device. A switch of some sort.
I already pointed this out in the thread, scroll down. The idea of a kill switch makes no sense. If the decision is made that some tech is dangerous it will be made by the owner or the government. In either case it will be a political/legal decision not a technical one. And you don’t need a kill switch for something that someone actively needs to pump resources into. All you need to do is turn it off.
there’s a whole lot of discussion around this already, going on for years now. an AI that was generally smarter than humans would probably be able to do things undetected by users.
it could also be operated by a malicious user. or escape its container by writing code.
Well aware. Now how does having James Bond Evil Villain-like destruction switch prevent it?
We have decided to run the thought experiment of a malicious AI is stuck in a box and wants to break out to take over. Ok, if you are going to assume this 1960s b movie plot is likely why are you solving the problem so badly?
As a side note I find it amusing that nerds have decided that intelligence gets you what you want in life with no other factors involved. Given that we should know more than anyone else that intelligence in our society is overrated.
