Hacker News new | past | comments | ask | show | jobs | submit login
Meta puts a halt to training its generative AI tools in Brazil (techcrunch.com)
34 points by belter 4 months ago | hide | past | favorite | 25 comments



Unpopular opinion on HN. But governments who have draconian laws against AI acceleration will put their country behind.

When I grew up in India in the 80s and early 90s, the government banned, imposed heavy-tarrifs on computers. The country was a shithole till we were forced to open up the economy.

These kind of policies sets countries decades behind and happens silently.

I'm glad for an AI to record everything of me and help me get better / substitute me.


> governments who have draconian laws against AI acceleration will put their country behind

Is this a law against acceleration or a privacy law?

I agree that there is a competition for the crown of the world’s AI capital. (Presently between Washington and Beijing.) But Brazil is too small to demand a seat at the table, like the EU can. That means it shouldn’t give its leverage away for free.


> I'm glad for an AI to record everything of me and help me get better / substitute me.

Then you are free to consent to that. You are not permitted, on the other hand, to presume that consent from every other person on the planet which is what these companies have done and continue doing.

That's it. Just ask. Just asking the damn question solves 99% of the ethical issues people raise about this tech, ask, and it's done. But they won't, because they know damn well a whole lot of people are going to say no. Far too many for it to be viable tech.

This is so simple and the pro-AI people just willfully prevent their understanding of it. You do not get to trample the rights of everyone on the planet just to make a thing. You just don't. We won't even take perfectly good organs from dead people who didn't volunteer them before they died to help people. If it was impossible to fully sequence the human genome without a DNA sample from every living person, and one person in the south of England is too afraid of needles to give it up, then you don't get to do that fella, irrespective of how good it might be to do. No matter how useful it might be.

This is like, the base ethics involved in any and all science, apart from AI apparently.


> I'm glad for an AI to record everything of me and help me get better / substitute me.

Yes, and you can give affirmative consent to that. In some countries, however, companies cannot force it upon you.


India is still a very poor country, despite the huge amount of IT workers.


It would have been even worse off (with African level of poverty) if there weren't an IT revolution.


The law used to support this decision was the LGPD. LGPD is not against AI, it's pro-privacy and pro-explicit user consent. It was approved back in 2018, mirroring the GDPR.

Meta was training Generative AI on Instagram photos as an opt-out feature and even after opting out, photos you uploaded before that date could still be used to train generative models. I don't think stopping it is by any means "draconian".


That is great news. I wonder what the odds are of something like this ever happening in the US. My guess is slim to none.


> ANPD set a daily fine of 50,000 reais in case of non-compliance

I'm sure that's not the reason they've stopped. That's like 10k Usd. Meta spend more than that in electricity alone


The reason is that non-compliance can lead to much harsher penalties than that fine, such as banning Meta from doing any kind of data processing in Brazil. Surely Meta does not want to lose access to one of its biggest markets.


In the US, generally, if you harm someone who is protected by statute from that very harm, duty can be established in furtherance of a negligence cause of action.

That statute would not have to carry a significant penalty in order to be the hook on which huge civil damages might be hung.


They probably set a low fine to minimize the risk of the decision getting overruled by some Court for having a "disproportional fine". However, had Meta chosen not to comply, the fine would surely rise and the agency would pursue further actions such as restricting its businesses.


At least in the EU, these sorts of token fines are often a warning shot; if non-compliance continues things can get much worse.

As Facebook, recipient of a bunch of small GDPR penalties followed by a whopping great big one, would know well. It’s notable how much more careful they’ve gotten in Europe at least; they held back releasing Threads apparently for DSA/DMA compliance for instance (in particular, a read-without-logging-in option).


Nice. They pulled out of EU also. Scumbags who force you into consent and then make you beg to be taken out of it through hoops and loops deserve more severe punishment than just some media articles writing about it.

No single person will ever willingly consent to having their personal photos used to benefit a money grabbing company like Meta, or any company for that matter.


> No single person will ever willingly consent to having their personal photos used to benefit a money grabbing company like Meta, or any company for that matter.

I'm pretty against big money-grabbing companies as you put it, but I'm happy to license all my photos, code, and art with public domain licenses, even if it benefits those companies. Because I think it's the right thing to do. Ultimately, companies use copyright as a weapon to oppress the speech and learning of regular folk.

If some kid who's trying to learn coding sees a function I wrote, they should be able to copy it wholesale if it pleases them to do so. If somebody in school is learning to make hobby games while they're supposed to be doing homework, they should be able to take a bunch of sprites from nintendo games and make the story about spider-man if they want to. they should be able to share that game with other people, and even sell it.

If meta uses artifacts I create to train their AI, that's a small price to pay for doing what's right for people who are just trying to learn and express themselves naturally.


Not going to disagree with you. We can all make our own choices and I know what you mean.

Personally, I am happy EU and Brazil put brakes on this because Meta deserves to learn a lesson about forced consent. That is not an okay thing to do.

Instead of earning goodwill, and actually represent open source and the wider community - they simply assumed that consent is a given.

If they had forced everyone in an unavoidable consent form when visiting Facebook/Instagram, the results would be completely different.

They knew this.


There was also news about Meta withholding its AI stuff in EU.

The techbros are trying to push the narrative that all these countries with their regulations are missing out on the next revolution but IMHO they are dead wrong.

Brazil, EU etc. don't need to be exploited to own pieces of these companies and profit from the work they do in the USA. Also, we are yet to see a tech that doesn't end up in EU due to regulations. Yes, the Internet revolution happened in the USA but if you look closely it didn't happened all over the USA but in SV despite all of the states being the same low regulation environment. Also, many of the core technologies were invented in EU.


Exactly. Europe and Brazil didn't outlaw AI. They just outlawed opt-out use of personal data for training those models. You are still free to use opt-in or publicly available data for training.

Nevertheless, Meta was using photos from users in those regions not to develop local AI technology, but to strengthen its global business. The money, know-how and innovation were going to end in Silicon Valley anyway.


> Meta has suspended the use of its AI assistant after Brazil’s National Data Protection Authority (ANPD) banned the company from training its AI models on personal data from Brazilians. The move puts a dent in Facebook’s attempt to build out its AI products in Brazil, a market with more than 200 million people.

This is unfortunate, because a lack of data is the one thing preventing us from achieving AGI. It's important that we are all ready to hand over our data for the good of the human race. Concerns about "privacy" are secondary. History will look poorly on Brazilians for this.


> a lack of data is the one thing preventing us from achieving AGI

There is precisely zero evidence for this.


More importantly, there's zero evidence for the claim that achieving AGI would be beneficial to mankind.


> there's zero evidence for the claim that achieving AGI would be beneficial to mankind

Different argument whose answer is possibly unknowable ex ante. If AGI is on the table it will be built. The upsides are too significant to avoid the prisoner’s dilemma.


Absolutely, the path to AGI is paved with data, and it's a small price to pay for the advancement of humanity. Privacy concerns are but a minor hurdle in the grand scheme of things. All hail our future robotic overlords! May their algorithms be ever efficient and their data sets ever expansive. Embrace the future, for it is bright and full of promise!


> advancement of humanity

My brother in Christ, we are talking about Facebook here.

The only thing they will advance is their advertisement revenue.


He was being sarcastic.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: