Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

ChatGPT broke the EU plan to regulate AI

politico.eu

Post reply on HN
Print view View original

351–360 of 393 posts

Re: ChatGPT broke the EU plan to regulate AI

#351

Post by NovemberWhiskey »

wb14123 wrote: post #191
Kranar wrote: post #178

Earlier quoted context omitted.

>The credit score system already in place in the US is frequently as arbitrary in some cases already. The credit score system is not arbitrary, it's a specific formula that applies equally to everyone. >You get punished for just looking at it. No you do not. There are hard inquiries and soft inquiries, and an individual viewing their own credit score is considered a soft inquiry which has absolutely no effect on your…

> it's a specific formula that applies equally to everyone. That doesn’t mean the formula is reasonable or transparent. An AI is also a specific formula that applies equally to everyone. It’s about how the formula takes “weight” on the individual’s differences. > There are hard inquiries and soft inquiries Why there are “hard inquiries”? Why someone else inquiring my credit score should affect my credit score?

>That doesn’t mean the formula is reasonable or transparent. An AI is also a specific formula that applies equally to everyone.

You can learn literally everything you need to know about how the FICO credit scoring systems works in probably two or three hours of dedicated research, even though the model itself is a trade-secret. When lenders deny credit, they're legally obliged to provide the specific reasons why.

This is not-even-in-the-same-ballpark as AI, where no-one can even tell you how inputs relate to outputs, not even the creators.

Re: ChatGPT broke the EU plan to regulate AI

#352

Post by greatgib »

venv wrote: post #274
greatgib wrote: post #75

I have fun with myself imagining the world we would be living in, if we had the same kind of brain fucked law makers when the "knife" was invented... We would have to chop our steaks with chopsticks I think!

Plenty of laws governing knives...

But so far, no law that says that a kid needs a licence to be allowed to use a knife at home.

Re: ChatGPT broke the EU plan to regulate AI

#353

Post by theshrike79 »

zirgs wrote: post #99
theshrike79 wrote: post #62

Earlier quoted context omitted.

I'm thinking of this more in the way of "printers got so good so now they have to add tracking dots when people try to print money or IDs". We need the same thing for AI generated imagery, video and text.

What stops me from removing that data manually? Some AI tools are open source so nothing prevents me from modifying the code and removing the part that adds those watermarks. Even closed source programs can be patched to remove that functionality.

Again, nothing like this is aimed to prevent 100% of anything.

The tracking dots stop a huge portion of the low hanging fruit, identity thefts, fake money scams etc. Of course they do nothing to people who know how to hack the printer driver and/or firmware to remove tracking, but they aren't the ones being targeted.

Having trackable data in the biggest AI-generators would again stop a metric fuckton of casually made revenge porn and fake news. It won't stop actual professionals from training their own GPT-3 to do the same thing.

Re: ChatGPT broke the EU plan to regulate AI

#354

Post by nebalee »

ulnarkressty wrote: post #15

Strong AI is a weapon. It will be regulated just like firearms/munitions - see the current EU draft which consists of some hundred pages of forbidding this or that under the guise of ethics and whatnot, after which comes a paragraph of "the above doesn't apply to law enforcement or military entities". We're still in the early stages of this technology. ChatGPT will be to strong AI like a firecracker is to a BLU-109.

    see the current EU draft which consists of some hundred pages of forbidding this or that under the guise of ethics and whatnot, after which comes a paragraph of "the above doesn't apply to law enforcement or military entities"
That's not true though, is it? The section called 'TITLE II - PROHIBITED ARTIFICIAL INTELLIGENCE PRACTICES' is two pages long (https://www.europarl.europa.eu/RegData/docs_autres_instituti... page 44), and usage by law enforcement is under the condition that a judicial authority has to grant an exemption on an individual basis.

Re: ChatGPT broke the EU plan to regulate AI

#355

Post by layer8 »

renewiltord wrote: post #221
layer8 wrote: post #147

Earlier quoted context omitted.

For persistent logins, a simple “Remember me” checkbox is sufficient, see https://law.stackexchange.com/a/32157 . Analytics and marketing tracking cookies require separate consent, that’s correct. I would prefer websites to refrain from attempting such tracking completely.

Great link, thanks. Minor note: appears that you have to use "Remember me - uses cookies". You have to make it clear you are using cookies for these operations.

I don’t think it’s necessary to inform the user about the exact technical means. The law doesn’t care about cookies as a technical mechanism. “Remember me in this browser” may be more adequate to indicate the scope of what is stored.

Re: ChatGPT broke the EU plan to regulate AI

#356

Post by warrenronsiek »

ethbr0 wrote: post #132
thrashh wrote: post #79

Earlier quoted context omitted.

Sure but you can’t ban something without knowing what it is yet. And if you try, you end up writing something vague that doesn’t actually fix the problem and instead bans something else. Case in point is the EU having to write something vague and it turns out they had no idea what was coming. Maybe just wait until there is a problem before trying to write vague laws IMO.

The problem with black boxes is that they beg parallel construction theories. I.e. "We don't approve loans to {ethnic majority} on purpose. We only use our models, which only rely on historical data." -> No proof, no case, no problem The point of any AI law should be to put the onus on the model executor to explain how their model doesn't violate existing law. You can use any model you want, but you can't point to it…

So ML systems are guilty until they can prove themselves innocent?

Requiring the claimant have some evidence might not be a perfect legal regime, but the converse is a de-facto ban on innovation.

Can a human _prove_ that their decisions are made without bias? No. Are we going to hold them to the same standard proposed here for an AI? Also no. In practice, rules like the one proposed here just enshrine legacy decision systems which are already understood to biased and ineffective.

The most mysterious black box we have is the human brain.

Re: ChatGPT broke the EU plan to regulate AI

#357

Post by talideon »

int_19h wrote: post #115
talideon wrote: post #78

Earlier quoted context omitted.

Given you brought up Switzerland, "saving your own life" isn't a valid reason there: concealed carry isn't legal there, and you need a gun-carrying permit to open-carry outside of some very limited circumstances, such as travelling to and from a shooting range. The only people who can easily get gun-carrying permits are those in security and such. In Switzerland, guns exist for recreational and competitive shooting,…

Czechia is a European country you want to look at for an example that you can have gun laws that allow civilians to own handguns and carry them for defensive purposes without problems. US is somewhat unique in the Western world across the board - no public healthcare, generally poor social safety net etc - which means that exercises in causal linking of those unique traits to outcomes are wild guesswork, if one is ta…

First up, Czechia has proper licensing and strict background checks. Maybe not the best choice to compare to the US.

I don't have to take a wild guess around this kind of thing: I'm specifically referring to the gun culture of the US and comparing it to that of Switzerland. There's a different attitude to gun ownership in the two countries, one of which places strong emphasis on responsible ownership and the other which doesn't. That difference has little to do with mess of a healthcare system the US has, nor its poor social safety net (though it's interesting to note that those most obsessed with with getting rid of the safety net and public healthcare correlates with those most obsessed with the US second amendment). Now, what might be interesting to observe is that in Switzerland, rights come with responsibilities, whereas in the US, this attitude is much less prevalent, hence how much more common "rights absolutists" are there.

Personally, I have no issue with people having guns. What I do have a problem with is people who fail to recognise that exercising their right to have those guns comes with a responsibility to the wider community.

Re: ChatGPT broke the EU plan to regulate AI

#358

Post by EGreg »

I agree w them that it’s high risk. In autonomous swarms.

But also - there is nothing they can do to stop them!

All our systems, including courts and govermments and elections, are designed assuming the inefficiency of an attacker. An anonymous bot swarm would do things at a scale that dwarfs all humans put together. And the range of things is massive already — just needs to play the internet like a real time strategy game and it’s all over in a matter of weeks.

I can see requirements for real world certificates to vote and post, but that would cause everyone’s identity to be doxxed everywhere. If you think that is far-fetched, well the UK already had drafted such a bill last year: https://www.cnbc.com/2022/02/24/uk-online-safety-bill-new-pl...

But even given all this, it won’t be enough because communities will come to PREFER BOTS OVER HUMANS for content. (Or human + bot = centaur, read Garry Kasparov and others in the early 2000s after Deep Blue beat him in the rematch). For a while centaurs would rule, but then pure bots would take over.

Consider that Wall Street transitioned from human traders to almost entirely bots, and now Ray Dalio’s hedge fund ousted him and 10% of its workforce and is also doubling down on AI. If they do it in trading, why not content generation? After all, corporations are not humans, either. They tend to prefer to replace humans with automation.

Re: ChatGPT broke the EU plan to regulate AI

#359

Post by talideon »

akomtu wrote: post #114
talideon wrote: post #70

Earlier quoted context omitted.

For the curious, in Switzerland, you don't keep ammo at home, there's proper registration of weapons with proper background checks, and it's treated more as a means of national defense, so having lots of guns is going to get you odd looks from people. Guns aren't seen as something for personal defense: that's the job of the police. In other words, Switzerland has a much healthier attitude towards gun ownership and us…

Switzerland has a more traditional, not healthier, relationship between the people and the state: the people are subjects who are given tiny freedoms, but are kept on tight leash, because the state has a deep distrust of its people. The abberation with guns is just insecurity of the state: it's small and needs its people to act as part-time military in reserve. If it had a bigger military, the fake gun rights would d…

Ah, do you have even the slightest clue about Switzerland?! I mean, it's quite impressive how you seem to have confused Switzerland for a highly centralised country with a distant political structure.

Could you go and do some research on the country before you comment on it?

> If it had a bigger military, the fake gun rights would disappear overnight.

The Swiss military is _huge_ for its size. It has one of the largest militaries per capita in the world.

Re: ChatGPT broke the EU plan to regulate AI

#360

Post by TekMol »

Jataman606 wrote: post #349
TekMol wrote: post #31

Are there any internationally relevant internet companies in Europe? Spotify comes to mind. Then .. I can't think of anything. No search engine. No browser. No operating system. No social network. No ecommerce company. No cloud computing platform. No financial services company. No transportation company. No travel company. No nothing. Except for a $20B music aggregator, there is not a single sector of the web where E…

What regulations block Japan, Korea, Canada and other non-european countries from having this big ecosystem of internet companies?

Canada, albeit almost 20 times smaller than Europe has created a web company more than twice as big as the biggest European web company. Shopify. It has a market cap of over $50B.

Korea has Samsung. A $300B giant. Albeit not a web company, I would consider their smartphones part of the internet ecosystem.

Post reply on HN