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

131–140 of 393 posts

Re: ChatGPT broke the EU plan to regulate AI

#131

Post by skissane »

corbulo wrote: post #103
hourago wrote: post #42

Earlier quoted context omitted.

> Ban social scoring, not the AI, the tool doesn't matter. Automated human classification without recourse nor transparency is what the EU is against. "You cannot get a loan because the AI says so" is not acceptable. > Would it be good if it was computed by hand on paper instead of through AI? Yes. As a judge can check what information was used and what algorithm applied to get to a conclusion. And it may declare tha…

>Automated human classification without recourse nor transparency is what the EU is against. "You cannot get a loan because the AI says so" is not acceptable. The credit score system already in place in the US is frequently as arbitrary in some cases already. You get punished for just looking at it. I don't think I need to preach to HN about that (crypto money and all that). Does Europe do it much differently? Not an…

> The credit score system already in place in the US is frequently as arbitrary in some cases already. You get punished for just looking at it. I don't think I need to preach to HN about that (crypto money and all that). Does Europe do it much differently?

Almost every country has some kind of credit reporting system, but many of them differ from the US in how they do it. In particular, in many countries, credit bureaus only hold negative information such as failing to pay bills or repay debts – which means if you keep up to date with all your bills and debts, your credit history will be blank. At which point, lenders will generally ask you to provide evidence of your employment history, salary, bank balance, assets, etc, and base their lending decision on that. They may well internally convert that all info into a "score", but every lender is going to have a different scoring system. Here's an article – https://www.businessinsider.com/credit-score-around-the-worl... – which discusses the systems used in different countries, including several European countries. There is no uniform approach to this in Europe – the UK and Germany are closer to the US in their approach, the Netherlands, Spain and France are closer to the "credit bureaus hold negative info only" system.

Re: ChatGPT broke the EU plan to regulate AI

#132

Post by ethbr0 »

thrashh wrote: post #79
bobthepanda wrote: post #14

Earlier quoted context omitted.

Control may be the wrong thing here. Every technology, at some point, will inevitably bring questions of legal liability and culpability, and AI has a large tendency to do so because of how applicable it is. It’s not the worst idea to try and get ahead of the problem by defining a legal framework. Examples of how a free-for-all has resulted in legal questions; * police and the judiciary in the US have come under fire…

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 its unexplainability as a free pass for ignoring protected classes.

And if you can't explain its results? Well, then you're opening yourself up to lawsuits.

Re: ChatGPT broke the EU plan to regulate AI

#133

Post by shagie »

xdennis wrote: post #68
user249 wrote: post #4

Good. Thanks to the EU I have to click "cookie question" every where I go. Thanks guys for wasting the limited time I have in life.

It's not the EU's fault that you have to click cookie banners. Those banners are only required if a website plans to do malicious things with the cookies. If they're used to track who's logged in, they are not required. They are more akin to the "Do not eat" warnings on silica packs... except on the internet everyone swallows.

> If they're used to track who's logged in, they are not required.

https://gdpr.eu/cookies/

    To comply with the regulations governing cookies under the GDPR and the ePrivacy Directive you must:

    Receive users’ consent before you use any cookies except strictly necessary cookies.
    Provide accurate and specific information about the data each cookie tracks and its purpose in plain language before consent is received.
    Document and store consent received from users.
    Allow users to access your service even if they refuse to allow the use of certain cookies
    Make it as easy for users to withdraw their consent as it was for them to give their consent in the first place.
If you want to save a person's login to make it easier for them to log in when they come back? That's not strictly necessary - consent is needed. If you save settings to a cookie - that's not strictly necessary - consent is needed. And then there's the "using a cookie to track a session to determine page bounce rate - even if it's not Google Analytics" - consent is needed.

And of course, consent is needed if you are using cookies for marketing.

Re: ChatGPT broke the EU plan to regulate AI

#134

Post by lillecarl »

vishal0123 wrote: post #93
dmix wrote: post #80

Earlier quoted context omitted.

The complete lack of design standards is what kills the utility. They should all have 3 buttons: "accept all" or "reject all" or "customize", dead simple. Every time it's a different design, different button text, different options. Usually rejecting = multi layers of options. A perfect example of good intentions making bad policy.

I have hard time believing it is not intentional. I think Europe could easily enforce consistent UI which could be automated through ad blocker, but now they know that most companies would rather stop serving Europe than not track users for ad.

I doubt "most companies" would want to lose out on one of the biggest economies on our planet. Some have built unsustainable businesses incompatible with privacy though.

I'm using uBlock and Consent-o-matic to remove as much tracking as possible already.

There's already the "do not track" header that noone respects.

Re: ChatGPT broke the EU plan to regulate AI

#135

Post by jareklupinski »

dmix wrote: post #80
user249 wrote: post #4

Good. Thanks to the EU I have to click "cookie question" every where I go. Thanks guys for wasting the limited time I have in life.

The complete lack of design standards is what kills the utility. They should all have 3 buttons: "accept all" or "reject all" or "customize", dead simple. Every time it's a different design, different button text, different options. Usually rejecting = multi layers of options. A perfect example of good intentions making bad policy.

> Every time it's a different design, different button text, different options

this is on purpose

if you could commit where the Deny button was to muscle memory, you would click it every time

Re: ChatGPT broke the EU plan to regulate AI

#136

Post by ilyt »

amelius wrote: post #86
felipelemos wrote: post #13

Earlier quoted context omitted.

In this case it's the opposite: they are trying to ban guns while not making murder a crime.

Well, it would make sense if they (at least temporarily) banned guns if they didn't know yet that they could be used for murder.

It's at "this thing shoots high speed metal projectiles at things far away in speed enough to go thru. We'd like to ban it before it becomes used for murders"

Re: ChatGPT broke the EU plan to regulate AI

#137

Post by ilyt »

ip26 wrote: post #17
aftbit wrote: post #11

Earlier quoted context omitted.

Yes, which is also a valid argument. Tools have no alignment. The people using them do. Thus, one should regulate the people, not the tools.

How about harmful tools with no practical uses to the general public? Mortars, tanks, ketamine, ICBM GPS units… You’re taking an absolutist position, so I’d like to explore it.

Well, tanks and mortars might've moved the scales on Emu War in favour of australians

Re: ChatGPT broke the EU plan to regulate AI

#138

Post by gopher_space »

px43 wrote: post #105
barrkel wrote: post #74

Earlier quoted context omitted.

How do you leap from banning black box AI decisions to thinking it means everything is allowed?

Haven't loan decisions always been black box? Sometimes they'll attribute decisions to credit scores, but those scoring systems have always been proprietary, and they won't even tell you what data your scoring is based on.

> Haven't loan decisions always been black box?

Not to the people making those decisions. That's brand new.

Re: ChatGPT broke the EU plan to regulate AI

#139

Post by ilyt »

influx wrote: post #65
ip26 wrote: post #39

Earlier quoted context omitted.

There is a tension between the need to level the playing field and the need to minimize the harm potential. For example, land mines, machine gun emplacements, and chemical weapons can be effective forms of home defense. An intruder would have access to the same equipment, but at least the playing field is level, right? However, an attacker would also be able to inflict astounding mass casualties in moments with this…

I appreciate the thought you've put into this, and I tend to agree. Ideally, a weaker person would be able to use a tool to provide self defense in the most limited way possible without causing harm or damage to anyone other than the attacker. Unfortunately the tools we have currently aren't great at that.

> Ideally, a weaker person would be able to use a tool to provide self defense in the most limited way possible without causing harm or damage to anyone other than the attacker.

Landmines in the garden it is then. Hard to use offensively.

Re: ChatGPT broke the EU plan to regulate AI

#140

Post by KyeRussell »

skissane wrote: post #131
corbulo wrote: post #103

Earlier quoted context omitted.

>Automated human classification without recourse nor transparency is what the EU is against. "You cannot get a loan because the AI says so" is not acceptable. The credit score system already in place in the US is frequently as arbitrary in some cases already. You get punished for just looking at it. I don't think I need to preach to HN about that (crypto money and all that). Does Europe do it much differently? Not an…

> The credit score system already in place in the US is frequently as arbitrary in some cases already. You get punished for just looking at it. I don't think I need to preach to HN about that (crypto money and all that). Does Europe do it much differently? Almost every country has some kind of credit reporting system, but many of them differ from the US in how they do it. In particular, in many countries, credit bure…

Yup. Until recently, Australia only had negative event reporting. Still, you’d regularly have kids that had been brainwashed by US-centric personal finance gurus (and American media more generally) asking how they could boost their “credit score”. Fully willing to sign up for and use credit cards as an 18-year-old student to juice some system they had been led to believe existed in the first place.
Post reply on HN