I have mixed feelings about this. Banning these technologies and then putting a few exceptions for the government isn’t an ideal solution to this problem. It feels like a power-grab. If you really want to for example tackle mass-surveillance, remove many of the cameras and reduce centralized control so that the police actually have to go and physically request recordings. But they are not doing that.
I haven’t mentioned ChatGPT because this AI act seems to extend beyond just ChatGPT and as the article mentions also sets rules for facial recognition and other similarly problematic technologies.
I’ve considered this as the totalitarian end result of AI.
That someone of x age, y status, flies in to Dubai, goes to z set of bars after checking in to a hotel, after b time if they haven’t gone to a restaurant, is c times likely to be found at d set of brothels, with e substances.
It’s all patterns, and it doesn’t matter if you participate or not, everyone else is.
It's clearly meant to invoke thoughts of Minority Report. There's not a court in the EU that will convict someone on a prediction that they will commit a crime. Made by computer model or otherwise. So it's stupid as hell to ban it. It's already banned.
And if the models are only used to know where to police, then my response is that that we've been using that for centuries now. It's called statistics, and can be as simple as bloody counting how many people lose their wallet in a busy street. Get some cops there, arrest the pick-pockets. That's a good thing, and it doesn't become "evil" just because somebody calls it "AI".
But this is just politicians piggy-backing on some popular fear-mongering about some new tech. It's unlikely to go anywhere. It may be rolled out again in a few months or years when these politicians need some attention again.
It's bad enough getting profiled by the likes of Equifax or some landlord app or bank redlining. It "might" be worse getting profiled as "crime-prone". By any organisation with power over your life, public or private.
Excellent news - more effort should be put into protecting intellectual property. Anything resembling original work, or training models against copyrighted work without permission, must be sanctioned. We cant have corporations steal our intellectual goods and then reselling them with different packaging.
It would appear that those guns are now turned against us so they can take our data and work and use it for free. Microsoft and openai are not training their models against their own products. Those are copyrighted.
But what we are told is that _our_ work is not copyrighted and they can just use it and sell it and profit from it. In return we will be “freed” of the burdens of our jobs and ip.
As I recall, the term for a pirate, operating legally and targeting government sanctioned targets was privateer. So maybe the term we need is privateering?
So, if you write a really good (free, since you don’t believe in intellectual property) book and I merely translate it into German, you don’t mind me selling it on Amazon and making money off of it? Ethically this is OK right? Since it’s not physical.
I don’t mean to come off as rude, I just find these “intellectual property doesn’t exist” takes ultimately very shallow and dismissive.
No, I get it, I’ve heard that argument and I see the logic behind it. It’s a “victimless” crime.
There are two issues I can think of that I would lead one to different conclusions. The first is that of consent - do I consent to my hard work being used without any reference to me and my effort? Would I have made the effort in the first place if there was no recognition?
The second is the following example: I quit my job to write an encyclopedia, something which is sure to benefit humanity. I toil day and night in order to one day sell it and get bread on the table. I finally publish it, but the next day another publisher copies it and sells it for cheaper (which they can afford, because they’re not paying me!), gobbling up my publisher’s market share and hence making me penniless, since the publisher has no money to pay me. Do I not deserve justice, considering the encyclopedia has sold a million copies and I am struggling to make a living?
The point is that the world is much too complex to draw a straight line between physical and intellectual property. A lot of us - and the vast majority on HN, I can imagine - work “intellectual” jobs in order to afford “physical” food.
If you respect the license. You cant steal open source code and just implement it in your products. Nor can you sell it modified or not without attribution and in some cases without open sourcing your own product.
Yeah counterfeit fashion products are not just illegal they also cause massive issues, both for original makers and consumers. It’s hilarious how all arguments against copyright as so blatantly easy to refute.
I'm not the parent poster, but I still wanted to reply.
In this hypothetical scenario, there's no copyright, so everything that isn't private is essentially public domain.
I've released my book into the public domain, you translated the book into German and also released it into the public domain. In addition, you suggest to the public that if they want to compensate you for your work (of translating the book to German), they can do so by purchasing a copy on Amazon. As long as you are honest about who wrote the book and what work you did (translating it), I'd have no problem.
Possibly, I've also indicated some way for people to compensate me for my work (even though they technically don't have to, they may want to do the moral thing and compensate me for the work I've put into something they enjoyed), for example I've also put it up for purchase on Amazon. It would then certainly be nice of you to also point your German readers to my Amazon entry in case they feel like giving something to me as well (or possibly even only to me).
This to me seems perfectly fine and ethical to me and I'd welcome that world.
But, continuing the hypothetical, why would you do that? Why would you spend the effort to compile all your knowledge in book form? Typesetting is hard, and you may want to use the expertise of other people for editorial review, printing and binding. Everything you pay them will be out of your own pocket, since the expected market value of the finished product is zero. On top of that, each one of those people may decide to publish your book as their own, and may have to if they want to recoup their own costs.
Taking a slightly wider view, how are you going to convince the bank to lend you a mortgage if your primary activity can only be remunerated via donations? In this world, who would still spend time on publishing books?
Before answering, I'd like to again remind you that we're talking about a world where the concept of intellectual property does not exist.
> your primary activity can only be remunerated via donations
I don't think the word "donations" really applies here, just like tips at a restaurant are not considered donations. Tips are not legally required, but they are heavily socially encouraged. If everyone agrees that creative works and knowledge should be shared and that we want more of them to exist, then it is expected, though not required that you give something back. Ideally proportional to the use you got out of it and what you have in the first place.
> On top of that, each one of those people may decide to publish your book as their own.
They can share the book all they want, but that doesn't mean they get to lie about who made the book ("as their own"), that should probably be illegal ;)
> But, continuing the hypothetical, why would you do that? Why would you spend the effort to compile all your knowledge in book form?
Maybe I hope to get compensated for it, maybe I like writing books, maybe I wrote the book for my own use and decided to share, maybe I personally value getting this knowledge out there for others to use which may even benefit me, maybe I feel grateful for the knowledge I received for free myself and want to give soemething back or build upon it or maybe I'm just obssessed with books :)
> Typesetting is hard, and you may want to use the expertise of other people for editorial review, printing and binding. Everything you pay them will be out of your own pocket, since the expected market value of the finished product is zero.
Probably if I only know about writing the book, it would make more sense to release a PDF, and maybe someone else will want to go through the trouble of actually making it into a printable book for some portion of the tips. Maybe they'd like to cooperate with me on the "official" release of the book for some share of the tips, after all anyone could possibly piggyback on their derivative work (or just get the PDF :)) so they may get more out of it by being part of the "original" "official" version of my book. Or maybe I can just pay them if it seems valuable to me.
> and may have to if they want to recoup their own costs
Or as above, they'll find it easier to recoup their costs by cooperating ;)
> Taking a slightly wider view, how are you going to convince the bank to lend you a mortgage if your primary activity can only be remunerated via donations?
As in our existing world, if the bank thinks I'll get enough back in tips, then it is a worthwhile investment. Also, in our hypothetical world, since everything is already available "for free", the best place to spend your budget for creative work is in works that don't already exist, making crowdsourcing a much more popular and rewarding option (I can get Eat Pray Love for free, but adverblessly's book is priceless since it will only exist if they get this money).
> In this world, who would still spend time on publishing books?
If we're talking about online publishing, I'm sure you could find plenty of people that will setup a website to host PDFs (or epub or mobi :)) for a 10% comission, guaranteeing the original artist gets at least 85% of the sale price (if it isn't going to the original artist, why even pay?). If w'ere talking about physical book publishing we'd probably see a similar dynamic. Alternatively, maybe you'll get your PDF from the original artist to tip them and then go to the professional printer for the physical book at printing cost +10% profit margin (at whichever book printer provides the best cost/quality for your needs).
The question is how this is really possible without a dramatic restriction of internet content accessible to citizens, i.e. national firewalls. OpenAI is accountable, but that’s only because they have something to lose. How does one reign in harmful services operating from outside the EU, who either can’t be reached or will not bother paying EU fines?
By pushing EU citizens to use services overseas, they can freely be spied on. So, enact strong privacy locally, then when looking at the next generation shift, make sure you can see exactly what your citizens are doing with it. “No laws broken”.
As always, the devil is in the details. They defined an AI system as software that uses any of:
(a) Machine learning approaches, including supervised, unsupervised and reinforcement learning, using a wide variety of methods including deep learning;
(b) Logic- and knowledge-based approaches, including knowledge representation, inductive (logic) programming, knowledge bases, inference and deductive engines, (symbolic) reasoning and expert systems;
(c) Statistical approaches, Bayesian estimation, search and optimization methods.
That could very reasonably be interpreted to mean all software. The act is mostly concerned with high risk applications. These are based on intended use, which is obviously tricky in the context of general purpose LLM’s. Most of the covered intended uses make sense, but in combination with their definition above it could mean that, for example, any developer of EdTech software in the EU could be regulated under the act.
More broadly, the act was clearly not written with general purpose foundation models in mind. For example, Article 10 requires that all training data be “relevant, representative, free of errors and complete.” That is pretty much impossible to comply with if you’re training on The Pile or similar.
> They defined an AI system as software that uses any of
I think you have genuine concerns, but no, you misunderstand.
You're not quoting the definition of an AI system. You're quoting Annex 1 (of the original draft), which defines “ARTIFICIAL INTELLIGENCE TECHNIQUES AND APPROACHES referred to in Article 3, point 1”. The definition of an AI system in said Article 3 point 1 has additional requirements.
Finally, to allay your fears, the draft legislation is proportionate: it focuses on “significant risk” and “foundation model[s]”, not every use of PyTorch.
That new definition still encompasses all software systems:
‘artificial intelligence system’ (AI system) means a machine-based system that is
designed to operate with varying levels of autonomy and that can, for explicit or
implicit objectives, generate outputs such as predictions, recommendations, or
decisions that influence physical or virtual environments
> Article 10 requires that all training data be "relevant, representative, free of errors and complete."
This is especially interesting to me with regard to something like ChatGPT. As we know, ChatGPT occasionally gives factually incorrect information. Does this mean that, in its current form, it would be illegal in EU? We know that Google is currently blocking access to Bard in EU. Will ChatGPT be forced to follow suit?
ChatGPT is great and I love it. It would be a shame if I'm not even allowed to use it _at my own risk_ just because it might be wrong about some things. This seems like a simplification, but it sounds like EU is allowing Perfect to be the enemy of Good.
It will be interesting to see what effects regulations like this will have on EU's access to emerging AI technology. We already saw Google block EU access to Bard.
"One thing that makes me skeptical about the EU long term is the way they seem to view regulation not as an unfortunate necessity, but as an exciting project."
> The main principles of EU law are equality before the law; legal certainty; predictable, clear and comprehensible legal provisions; and transparent legislation. 2 These principles can be upheld only if EU law has the same meaning in all language versions.
“Open to interpretation” has nothing to do with the quote above. It doesn’t matter what the US has and what it does not. We’re not discussing the US here.
> It doesn’t matter what US has and what it does not. We’re not discussing US here.
When you claim the EU “has no idea” apropos nothing—simply for having a mechanism for interpreting statues in court—it is entirely reasonable to question your double standard.
All major legal jurisdictions have courts to resolve things...
“Apropos nothing” added by yourself. Look - it doesn’t matter how the law is interpreted in the US because we’re not discussing the US here and the US is totally irrelevant to this discussion.
This AI act, together with the CRA, are written by the people who, vast majority of them, never worked with software. Yet, they intend to regulate software. This will go wrong.
It would be one thing if the heavy-handed, often premature, EU regulations were in some way intended to foster production of its own technology companies to compete at large scale in its own market. But we all know that doesn't happen. Any European with an entrepreneurial spirit is all but forced to go start their company in USA. It seems like anything that appears potentially useful but has undefined or unanswered legal questions is just banned before anyone can figure out what the actual problems are. Sigh.
I guess once we get to 1M characters in the context window, 90% of it would be preloaded by all existing laws in order to avoid any hint of breaking the laws. In areas where two or more laws contradict each other, LLMs will behave like insane persons.
German speaking. I'm seeing some defenders of this (and other EU IT regulation attempts). I used to think the same way in times of GDPR:
1. How do we as tech geeks believe that we're somewhat special and won't ever get regulated given that Software is eating the world.
2. I trust the EU and believe we should push things forward.
3. If they get it wrong, they will monitor and fix it (boy this is so cute, me 6 years ago).
4. We cannot allow anymore having BigTech invade all our industries, spy on and buy/copycat our local competition killing our economy long-term.
Boy was I wrong! All we have are cookie-banners everywhere, legal insecurity and legal battles initiated by third-class law firms crawling the web for the quick buck, understaffed courts that are overwhelmed with cases and BigTech with their big law departments having almost no problem with all that; in fact it's helping them a lot.
In addition, EU is already showing attempts to put even more control over its citizens, like chat control which is essentially the pipe dream of old Stasi majors and don't forget that EU people started working on a governmentally controlled DNS server, currently as an opt-in, but it's a matter of time when they will start requiring orgs and citizens to use it.
5 years from now, we will see that we didn't catch up with BigTech, we will just have local "innovators" who essentially resell Google/AWS/MS/OpenAI stuff. They will put more and more regulation pressure on their citizens "for the good" and we're just a few bad decisions away from the perfect politically controlled digital dystopia.
They don't know what they're doing, and they're even excited doing it.
52 comments
[ 3.1 ms ] story [ 120 ms ] threadI haven’t mentioned ChatGPT because this AI act seems to extend beyond just ChatGPT and as the article mentions also sets rules for facial recognition and other similarly problematic technologies.
They watched Minority Report and thought "that should get some attention." Ridiculous.
“Dubai Police using AI to predict crime” (https://www.arabianbusiness.com/culture-society/dubai-police...)
That someone of x age, y status, flies in to Dubai, goes to z set of bars after checking in to a hotel, after b time if they haven’t gone to a restaurant, is c times likely to be found at d set of brothels, with e substances.
It’s all patterns, and it doesn’t matter if you participate or not, everyone else is.
And if the models are only used to know where to police, then my response is that that we've been using that for centuries now. It's called statistics, and can be as simple as bloody counting how many people lose their wallet in a busy street. Get some cops there, arrest the pick-pockets. That's a good thing, and it doesn't become "evil" just because somebody calls it "AI".
But this is just politicians piggy-backing on some popular fear-mongering about some new tech. It's unlikely to go anywhere. It may be rolled out again in a few months or years when these politicians need some attention again.
- Police stalking (legal in the Netherlands, practiced but denied in many other countries).
- Police asking neighbors questions, thereby ruining your reputation.
- Database sharing with e.g. the U.S., leading to further scrutiny (or denial of entry) abroad.
- Not getting a government job.
You are entitled only to what you created.
But what we are told is that _our_ work is not copyrighted and they can just use it and sell it and profit from it. In return we will be “freed” of the burdens of our jobs and ip.
In effect this is reverse piracy.
I don’t mean to come off as rude, I just find these “intellectual property doesn’t exist” takes ultimately very shallow and dismissive.
I never will understand why this is so hard for people to grasp. I’m not entitled to anything else.
https://youtu.be/IeTybKL1pM4
There are two issues I can think of that I would lead one to different conclusions. The first is that of consent - do I consent to my hard work being used without any reference to me and my effort? Would I have made the effort in the first place if there was no recognition? The second is the following example: I quit my job to write an encyclopedia, something which is sure to benefit humanity. I toil day and night in order to one day sell it and get bread on the table. I finally publish it, but the next day another publisher copies it and sells it for cheaper (which they can afford, because they’re not paying me!), gobbling up my publisher’s market share and hence making me penniless, since the publisher has no money to pay me. Do I not deserve justice, considering the encyclopedia has sold a million copies and I am struggling to make a living?
The point is that the world is much too complex to draw a straight line between physical and intellectual property. A lot of us - and the vast majority on HN, I can imagine - work “intellectual” jobs in order to afford “physical” food.
/s
In this hypothetical scenario, there's no copyright, so everything that isn't private is essentially public domain.
I've released my book into the public domain, you translated the book into German and also released it into the public domain. In addition, you suggest to the public that if they want to compensate you for your work (of translating the book to German), they can do so by purchasing a copy on Amazon. As long as you are honest about who wrote the book and what work you did (translating it), I'd have no problem.
Possibly, I've also indicated some way for people to compensate me for my work (even though they technically don't have to, they may want to do the moral thing and compensate me for the work I've put into something they enjoyed), for example I've also put it up for purchase on Amazon. It would then certainly be nice of you to also point your German readers to my Amazon entry in case they feel like giving something to me as well (or possibly even only to me).
This to me seems perfectly fine and ethical to me and I'd welcome that world.
But, continuing the hypothetical, why would you do that? Why would you spend the effort to compile all your knowledge in book form? Typesetting is hard, and you may want to use the expertise of other people for editorial review, printing and binding. Everything you pay them will be out of your own pocket, since the expected market value of the finished product is zero. On top of that, each one of those people may decide to publish your book as their own, and may have to if they want to recoup their own costs.
Taking a slightly wider view, how are you going to convince the bank to lend you a mortgage if your primary activity can only be remunerated via donations? In this world, who would still spend time on publishing books?
> your primary activity can only be remunerated via donations
I don't think the word "donations" really applies here, just like tips at a restaurant are not considered donations. Tips are not legally required, but they are heavily socially encouraged. If everyone agrees that creative works and knowledge should be shared and that we want more of them to exist, then it is expected, though not required that you give something back. Ideally proportional to the use you got out of it and what you have in the first place.
> On top of that, each one of those people may decide to publish your book as their own.
They can share the book all they want, but that doesn't mean they get to lie about who made the book ("as their own"), that should probably be illegal ;)
> But, continuing the hypothetical, why would you do that? Why would you spend the effort to compile all your knowledge in book form?
Maybe I hope to get compensated for it, maybe I like writing books, maybe I wrote the book for my own use and decided to share, maybe I personally value getting this knowledge out there for others to use which may even benefit me, maybe I feel grateful for the knowledge I received for free myself and want to give soemething back or build upon it or maybe I'm just obssessed with books :)
> Typesetting is hard, and you may want to use the expertise of other people for editorial review, printing and binding. Everything you pay them will be out of your own pocket, since the expected market value of the finished product is zero.
Probably if I only know about writing the book, it would make more sense to release a PDF, and maybe someone else will want to go through the trouble of actually making it into a printable book for some portion of the tips. Maybe they'd like to cooperate with me on the "official" release of the book for some share of the tips, after all anyone could possibly piggyback on their derivative work (or just get the PDF :)) so they may get more out of it by being part of the "original" "official" version of my book. Or maybe I can just pay them if it seems valuable to me.
> and may have to if they want to recoup their own costs
Or as above, they'll find it easier to recoup their costs by cooperating ;)
> Taking a slightly wider view, how are you going to convince the bank to lend you a mortgage if your primary activity can only be remunerated via donations?
As in our existing world, if the bank thinks I'll get enough back in tips, then it is a worthwhile investment. Also, in our hypothetical world, since everything is already available "for free", the best place to spend your budget for creative work is in works that don't already exist, making crowdsourcing a much more popular and rewarding option (I can get Eat Pray Love for free, but adverblessly's book is priceless since it will only exist if they get this money).
> In this world, who would still spend time on publishing books?
If we're talking about online publishing, I'm sure you could find plenty of people that will setup a website to host PDFs (or epub or mobi :)) for a 10% comission, guaranteeing the original artist gets at least 85% of the sale price (if it isn't going to the original artist, why even pay?). If w'ere talking about physical book publishing we'd probably see a similar dynamic. Alternatively, maybe you'll get your PDF from the original artist to tip them and then go to the professional printer for the physical book at printing cost +10% profit margin (at whichever book printer provides the best cost/quality for your needs).
Yes it is.
Do you know how the 5/14 eyes program works?
By pushing EU citizens to use services overseas, they can freely be spied on. So, enact strong privacy locally, then when looking at the next generation shift, make sure you can see exactly what your citizens are doing with it. “No laws broken”.
That could very reasonably be interpreted to mean all software. The act is mostly concerned with high risk applications. These are based on intended use, which is obviously tricky in the context of general purpose LLM’s. Most of the covered intended uses make sense, but in combination with their definition above it could mean that, for example, any developer of EdTech software in the EU could be regulated under the act. More broadly, the act was clearly not written with general purpose foundation models in mind. For example, Article 10 requires that all training data be “relevant, representative, free of errors and complete.” That is pretty much impossible to comply with if you’re training on The Pile or similar.
I think you have genuine concerns, but no, you misunderstand.
You're not quoting the definition of an AI system. You're quoting Annex 1 (of the original draft), which defines “ARTIFICIAL INTELLIGENCE TECHNIQUES AND APPROACHES referred to in Article 3, point 1”. The definition of an AI system in said Article 3 point 1 has additional requirements.
Moreover, the news here is a revised draft has been made. You can see the revised definition in Article 3 on page 137. https://www.europarl.europa.eu/meetdocs/2014_2019/plmrep/COM...
Finally, to allay your fears, the draft legislation is proportionate: it focuses on “significant risk” and “foundation model[s]”, not every use of PyTorch.
‘artificial intelligence system’ (AI system) means a machine-based system that is designed to operate with varying levels of autonomy and that can, for explicit or implicit objectives, generate outputs such as predictions, recommendations, or decisions that influence physical or virtual environments
This is especially interesting to me with regard to something like ChatGPT. As we know, ChatGPT occasionally gives factually incorrect information. Does this mean that, in its current form, it would be illegal in EU? We know that Google is currently blocking access to Bard in EU. Will ChatGPT be forced to follow suit?
ChatGPT is great and I love it. It would be a shame if I'm not even allowed to use it _at my own risk_ just because it might be wrong about some things. This seems like a simplification, but it sounds like EU is allowing Perfect to be the enemy of Good.
Obvious hyperbole, but:
There is no such thing as "at your own risk" in the EU. It's a Nanny State. Brussels knows best what's good for you.
> The main principles of EU law are equality before the law; legal certainty; predictable, clear and comprehensible legal provisions; and transparent legislation. 2 These principles can be upheld only if EU law has the same meaning in all language versions.
“Open to interpretation” has nothing to do with the quote above. It doesn’t matter what the US has and what it does not. We’re not discussing the US here.
When you claim the EU “has no idea” apropos nothing—simply for having a mechanism for interpreting statues in court—it is entirely reasonable to question your double standard.
All major legal jurisdictions have courts to resolve things...
This AI act, together with the CRA, are written by the people who, vast majority of them, never worked with software. Yet, they intend to regulate software. This will go wrong.
The special AI WG started this work years ago. Its probably just a few years younger than GDPR.
The gdpr-style 6% fine has been known for 2-3 years.
1. How do we as tech geeks believe that we're somewhat special and won't ever get regulated given that Software is eating the world.
2. I trust the EU and believe we should push things forward.
3. If they get it wrong, they will monitor and fix it (boy this is so cute, me 6 years ago).
4. We cannot allow anymore having BigTech invade all our industries, spy on and buy/copycat our local competition killing our economy long-term.
Boy was I wrong! All we have are cookie-banners everywhere, legal insecurity and legal battles initiated by third-class law firms crawling the web for the quick buck, understaffed courts that are overwhelmed with cases and BigTech with their big law departments having almost no problem with all that; in fact it's helping them a lot.
In addition, EU is already showing attempts to put even more control over its citizens, like chat control which is essentially the pipe dream of old Stasi majors and don't forget that EU people started working on a governmentally controlled DNS server, currently as an opt-in, but it's a matter of time when they will start requiring orgs and citizens to use it.
5 years from now, we will see that we didn't catch up with BigTech, we will just have local "innovators" who essentially resell Google/AWS/MS/OpenAI stuff. They will put more and more regulation pressure on their citizens "for the good" and we're just a few bad decisions away from the perfect politically controlled digital dystopia.
They don't know what they're doing, and they're even excited doing it.