Meta confirms it trains its AI on any image you ask Ray-Ban Meta AI to analyze | Hacker News Reader

Meta confirms it trains its AI on any image you ask Ray-Ban Meta AI to analyze

techcrunch.com

51 pointsmikhael18 comments

Gadgets were more fun before everything felt like a data miner. I used to buy all sorts of tech junk 20 years ago like PDAs, GPS, MP3 players, fitness trackers, etc. but the primary purpose of everything now feels like sending all my data to ad companies and locking me into a subscription. If something has a camera, microphone, GPS, or just wifi now my impulse is that it’s creepy or something to be weary of rather than useful.

I remember this time, some of the coolest non spy devices I remember were:

1. Sony Clie PDA.

2. Creative Nomad MP3 Player.

3. Most of my game consoles, Gameboys. I especially loved the Super Scope, I couldn't believe how cool that was.

4. Pretty much all my Linux devices until this day.

Having a w*k with my Apple Watch first the first time was unnerving

Show companies that your value of privacy trumps the value they get from the data. For example, how much would you pay for a version of YouTube where they don't log your views at all? Would you pay $30/month for it? I think today companies believe you don't really value it. You just give it lip service. I suspect if we show (financially) how much we value it, we'll see them respond.

For one, it's impossible to do what you suggest when no one offers such a thing. You can throw money at your screen, but it doesn't change the fact that there is no amount of money you can pay google to not track you.

Second, the data they extract from you is far more valuable than what anyone would pay. Besides that, there is absolutely nothing preventing them from taking your money and tracking you anyway. Even if they don't at first, a corporation cannot resist the 'free' revenue gained by tracking you and selling that data.

This is simply not something that the market can correct. It must be a legal mechanism enforced by the government.

They’ll want to get that publicly stated early and forgotten.

Stupid question but is this wise from a business standpoint? let's say child pornography or some national security stuff slips into training data.

This stuff is already part of the training data - Laion for example (which almost everyone uses as training data) has a lot of csam in it (allegedly I never checked)

This has already happened with ChatGPT.

how does this work in a 2-party consent state, like California? seems illegal.

I mean... Yes? If you ask a cloud hosted AI to identify an image, that image goes to the AI. Presumably if I am an active user of a service to identify objects in an image I want that service to correctly identify objects, which means if it incorrectly IDs something I upload today, I want it to be better at its job the next time I try.

In other news, voice transcription from Google trains itself on your voice. Oh and Google search trains itself on your searches. (Snark: assuming they still care about search quality!)

And a user could very reasonably expect all of your examples to be read operations and be surprised and unhappy to find that their data was logged.

All these companies are data ingestion machines. This is not news.

This one’s a bit worse than meta’s usual sins: enabling political ads to manipulate the user is one thing. But enabling naughty people to generate naughty pictures of innocent bystanders because they happened to be in the field of view of an idiot talking to their glasses is a whole different level. Would be surprised if this is legal.

Hardly worse than their usual sins considering the case of Meta and Myanmar.

well yeah.

the quest does interior room mapping of all spaces with every available sensor. user doesn't have to be actively engaged so long as the device is still on and logged in, they've already waived any rights, this is why sleep vs powering off is a such a konnundrum and the device will try to stay in standby mode at all costs.

not sure were this data goes tbh. so far only seems to be internal.

Of course it does. Any product that you use, which has camera, is constantly feeding images to Meta's AI to train on. That's why a long time back I stopped using Oculus devices as well.

Where is it stated they sent or can sent Meta Quest pass-through images to Meta? Persistent storage of Rrom wireframe mapping for positioning I get and don't find a big issue, I thought that was on-device though.

oh lol, you replied the same thing while i was typing and hadn't refreshed. heh.

Reply on news.ycombinator.com