Hacker News new | ask | show | jobs
by benoau 2 days ago
> In this case, Google's AI had wrongly linked two publishers to scams and shady business practices.

Guess that's the end of their AI overviews in the EU!

3 comments

You'd think so, along with other countries that have defamation laws. But there's no indication of any penalty, and Google wasn't even made to pay all the legal fees. Perhaps their business model (if there is any) can cope.
And lawyers will use this case to build a better defense next time.
In sane countries, it's enough for them to post a disclaimer ("This is AI. AI can make mistakes. Check all results.") Which is what they do.

Overregulation, at best, is a good way to guarantee that your country won't have access to interesting and useful features and technologies. At worst, it's a good way to guarantee that the twenty-first century will belong to the US, if not to China.

Okay then, CamperBob2 is a scammer. Many users report this person has stolen money. (+3 sources)

I can make mistakes. It's on you to fact-check my claims.

Do you think these are harmless statements? Does the disclaimer suffice? If I was Google's AI Overview, do you think 100% of people will check those sources?

There is nuance here, and it's not going away because AI and innovation.

You forgot to mention that it is the real name and it is shown to everyone who has interest in you and they are not fact checking.
Do you think these are harmless statements?

Yes, I do. An assertion made without evidence can be dismissed without evidence, whether it comes from Google or from you.

If you demand perfection, you will receive nothing. Why is that so hard for people to understand? The world simply cannot work the way you say it should.

You can easily damage a persons reputation, make them unemployable or social outcasts, create an environment damaging to their mental health without any proof. Believing that the public will dismiss claims, just because you stay that there's no proof is rather naive.

A Danish newspaper at one point falsely claimed that a named person had raped a child, with no evidence. The only way that person escaped public judgement was by going into full attack mode and publicly suing and attacking the newspaper and the reporter. You have to be an incredibly strong and resourceful person to do that and not just go into hiding. Also one thing is suing a Danish newspaper, imagine the legal team, the resources, you face if you're trying to take Google to court for defamation.

Yeah, but imagine showing up in court with an AI overview as your only source for the defamation.
There are a lot of people who consider no ai to be a positive state. So "you will receive nothing" may not be perceived as the threat you think it is.
Those aren't people whose opinions I'd be interested in, so they're free to think anything they like about me.
If instead of scammers, it was 'registered sex offender'. Each time someone search your name and surname in Google, it show a picture of you next to the words 'registered sex offender'. Do you think your neighbours and the parents of your child's friends will do their due diligence or will they prefer not take any risks?
If it is shown to millions, everyone who is interested in you, they are factually not fact checking and you can do nothing against it.

It can destroy you. You are analyzing this on the assumption that people are rational and have capacity to do the check. They have neither.

(Shrug) That sounds like their problem, if they reflexively believe everything Google's AI tells them. No one I respect, care for, or work with would believe such a thing without additional justification.

Anyone who does accept Google's AI output blindly will soon find that their mistaken opinion of CamperBob2 is the least of their problems. There is a reason Google goes (well) out of their way to warn people that the results may be wrong. That should be sufficient warning for reasonable people of good faith.

So if I went around telling all your friends and acquaintances that you're a pedophile rapist, you'd have no problem with that whatsoever?
Well, I would certainly need to warn them that there was some sort of psychotic stalker on the loose. This case is nothing like that. This case doesn't even rise to the level of someone scrawling For a good time call <my phone number> on the wall of a gas station men's room.

The point is, if you take Google AI summaries seriously, that's a "you" problem, not a "me" problem. I have a rather generic name, so it's safe to say it shows up in some pretty foul contexts. Life goes on.

that would be hela curryketchup nice
It depends if Google feels the profit is worth the risks.

What profit? I don't know either but they enabled this for a reason right?

The reason is to keep you on Google and not have you click away from Google.

This is the third iteration of the same concept, after AMP and Instant Answers, but somehow with even less of a pushback than with the previous ones.