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

Facebook is researching AI systems that see, hear, remember everything you do

theverge.com

Post reply on HN
Print view View original

121–130 of 144 posts

Re: Facebook is researching AI systems that see, hear, remember everything you do

#121

Post by dang »

thr0wawayf00 wrote: post #83
pesenti wrote: post #61

Earlier quoted context omitted.

I encourage you to read the details of the news, in particular how the data was collected with full consent from all (something that most research datasets don’t do well). It doesn’t answer your question directly but gives you an idea on how we think about these issues. AI-powered wearables will have stringent privacy guarantees, both for the wearer and the people around.

HN user asks a simple and direct question to Facebook's VP of AI: > What are you all doing to allow those of us that don’t want to be a part of your utopia to opt out? Answer from Facebook VP of AI: > I encourage you to read the details of the news that doesn't actually answer your question but gives you an idea of how we think I'm glad to see that even on a industry-specific forum like Hacker News, we can't get stra…

Please don't attack people like this when they post to HN, regardless of how you feel about their employer. You may not owe $BigCo better, but you owe this community better if you're participating in it. When something remains unanswered, it's enough to ask substantive questions respectfully.

We don't want HN to be a hostile place that tars-and-feathers people who show up to explain another side of a story—especially not on a topic they know about. For most of us, our work is what we know the most about. Therefore, people showing up to discuss something related to their work are among the highest-value contributors HN can have. We don't want to discentivize that, and attacking them for it goes directly against the mandate of the site. You don't have to agree, obviously, but you do have to stick to the site guidelines. (Btw, that also means that you shouldn't be posting generic-indignant rants here.)

https://news.ycombinator.com/newsguidelines.html

https://hn.algolia.com/?query=disincent%20by:dang&dateRange=...

Edit: I just noticed that you got much nicer later in the thread. That's way better—thanks.

Re: Facebook is researching AI systems that see, hear, remember everything you do

#122

Post by wonderwonder »

pesenti wrote: post #106
thr0wawayf00 wrote: post #96

Earlier quoted context omitted.

Great! How do I make sure that nothing about me ever passes through a Facebook server/ETL job/datalake/neural net/whatever? Can that even be guaranteed? Is Facebook generating a shadow profile on me or using data I've generated either myself or through friends that have taken pictures/videos/etc of me, and if maybe, how do I opt out? Facebook just released Ray Ban smart glasses that don't look any different from stan…

No, that’s not a reasonable assumption. Photos or videos of you are likely on Facebook’s servers, uploaded by friends or others when you are in public. What I can assure you though is that if you are not a user, we are not recognizing you in these pictures and videos. We won’t do that without explicit consent from users.

Thanks for answering questions. In regards to this. If I have a Facebook account, and then close it, is my existing data in Facebook's systems erased including image recognition / tracking based preferences, etc?

Re: Facebook is researching AI systems that see, hear, remember everything you do

#123

Post by mortehu »

still_grokking wrote: post #109
pesenti wrote: post #106

Earlier quoted context omitted.

No, that’s not a reasonable assumption. Photos or videos of you are likely on Facebook’s servers, uploaded by friends or others when you are in public. What I can assure you though is that if you are not a user, we are not recognizing you in these pictures and videos. We won’t do that without explicit consent from users.

What does "we are not recognizing you" mean? Of course you can't "recognize" someone you don't know, so what do you try to say here? Do you gather data about those "unknown" persons. Do you try to match information about "unknown" persons form different sources?

> Do you gather data about those "unknown" persons

This part is easy to answer: if a user uploads a photo of you and there are other people in the photo or a recognizable location, then that's data about you. Even if it's just stored as a photo now, it'll take a neural network a millisecond to turn that into feature vectors that can be used to build a knowledge graph. This goes for Dropbox as well as Facebook.

Re: Facebook is researching AI systems that see, hear, remember everything you do

#124

Post by stereoradonc »

pesenti wrote: post #39

I support Facebook AI. The actual news is that we are releasing an extensive egocentric dataset with associated tasks for the research community. Collecting such dataset in a private and responsible way is a significant challenge. Happy to answer any questions. https://ai.facebook.com/blog/teaching-ai-to-perceive-the-wor...

I support Facebook shutting down this program.

Re: Facebook is researching AI systems that see, hear, remember everything you do

#125

Post by zwaps »

I suggest everyone to carefully read the replies of the Facebook employees in this thread.

These data are being released in hopes of furthering the developments of these techniques - which ultimately benefits a future use by Facebook.

In turn, people like "presenti" have pointed out several times that the data were collected in Facebook offices or other places where all participants consent. The implication is clear: While the present data are legal, they would not be if shot in the "wild".

In many jurisdictions, there is no legal use case for these methods. It is literally not possible to have a widely-used implementation of these glasses that would not violate the privacy rights of "bystanders".

Of course, Facebook will work hard to make it legal, on all fronts. Whether laws need to be changed or, as pointed out by presenti, some innovation may arise to make you at least an anonymous blob, separate from name and address. And if you do not have an account - see in this thread - what is there to worry about? Well, even if the association would only be indirect, any data collected from you (with increasing sophistication) would be part of Facebook and its models. Even as a group of anonymous users, you would become less and less anonymous and increasingly more explainable.

If you are so far unconvinced that this matters, please be also aware that the only thing standing between businesses and the extraction of the entire surplus (or consumer welfare) is asymmetry of information. As soon as your needs and wants are sufficiently modeled, companies will use these data to maximize the profit from all their interactions with you. You will, in a sense, pay monopoly prices without facing monopolies.

Again, read the replies carefully. See how legal constraints are certainly something that is "to be solved in the future". However, whether you should ever be afforded privacy or anonymity in the face of facebook's algorithm is implicitly answered, with (frankly) a worrying amount of arrogance and dismissal: no, you and everything about you should no longer be any unexplained variation for facebook. One way or another, facebook plans to uncover you, and there is no negotiating this point.

You must now realize that Facebook is not interested in your name and birthdate, they are interested in being able to predict everything about you without these personal data. Faeebook wants a model, one that is fundamentally opposed to your welfare. And this objective will be realized. Take heed of this, better now than later.

Re: Facebook is researching AI systems that see, hear, remember everything you do

#126

Post by zwaps »

still_grokking wrote: post #110
thr0wawayf00 wrote: post #108

Earlier quoted context omitted.

Thanks Jerome, I really do appreciate your answer and I know I'm probably not the most fun person to talk to. I do have a follow-up on this if you'll entertain me: why is it that Facebook stores content of people that it doesn't recognize? Granted, I'm but a lowly web developer, but it seems like creating business logic that automatically removes content with people that aren't Facebook users would be pretty straight…

> Recognizing people in images is only one of many types of data that Facebook gleans from that content, and I don't want anything involving me being processed in any way by that company, whatsoever. Which is a basic right under GDPR. No private entity is allowed to store or process data about you for their purposes without your consent. Did FB just said they are doing it though?

Of course. FB openly ignores the GDPR and will be happy to pay a fine if the EU manages to ever show enough courage to levy one of significance.

Facebook trains its models on your data, in particular your pictures and (in the future) your video - your behavior, voice, face, etc. But feat not, your name will not flow directly in the algorithms (until you sign up an account - and the rest of your data is already there!)

And as presenti openly (and quite arrogantly) stated just one post up: There is no conceivable way for you to ever do anything about it, except if FB stops operations of this kind. Which they can't, of course, because it is the very basis of their business model.

Re: Facebook is researching AI systems that see, hear, remember everything you do

#127

Post by Osiris »

zwaps wrote: post #126
still_grokking wrote: post #110

Earlier quoted context omitted.

> Recognizing people in images is only one of many types of data that Facebook gleans from that content, and I don't want anything involving me being processed in any way by that company, whatsoever. Which is a basic right under GDPR. No private entity is allowed to store or process data about you for their purposes without your consent. Did FB just said they are doing it though?

Of course. FB openly ignores the GDPR and will be happy to pay a fine if the EU manages to ever show enough courage to levy one of significance. Facebook trains its models on your data, in particular your pictures and (in the future) your video - your behavior, voice, face, etc. But feat not, your name will not flow directly in the algorithms (until you sign up an account - and the rest of your data is already there!…

I'm not a lawyer but my understanding in the United States is that a person in a public place has no legal expectation of privacy. If you are in public and someone takes a picture in which your face is visible and then upload that to Facebook, I wouldn't expect that to be a violation of my privacy nor would I expect Facebook to allow an unknown person to request the picture's removal.

Re: Facebook is researching AI systems that see, hear, remember everything you do

#128

Post by 6gvONxR4sf7o »

pesenti wrote: post #113
jjulius wrote: post #95

Earlier quoted context omitted.

>I encourage you to read the details of the news, in particular how the data was collected with full consent from all (something that most research datasets don’t do well). I don't care to hear about the consent you received for this controlled study. I care to hear about how I/we can opt-out from something like this when it is eventually deployed in the wild en masse as part of an actual product. I'm pretty sure tha…

It's hard to elaborate on something that's not developed yet. Here are the principles we are using when developing these products: https://about.facebook.com/realitylabs/responsible-innovatio... In practice, it will mean that some things you will know about, but don't necessarily give consent (i.e., someone taking a picture or a video of you), while other will likely require consent (i.e., recognizing you in these pi…

> In practice, it will mean that some things you will know about, but don't necessarily give consent (i.e., someone taking a picture or a video of you)

This makes me and so many others upset. I hate that people can upload photos of me to social media without my consent. And with the new glasses, it’s disingenuous to even say I’ll even know it’s happening. The little light on it is a bad solution.

Re: Facebook is researching AI systems that see, hear, remember everything you do

#129

Post by still_grokking »

Osiris wrote: post #127
zwaps wrote: post #126

Earlier quoted context omitted.

Of course. FB openly ignores the GDPR and will be happy to pay a fine if the EU manages to ever show enough courage to levy one of significance. Facebook trains its models on your data, in particular your pictures and (in the future) your video - your behavior, voice, face, etc. But feat not, your name will not flow directly in the algorithms (until you sign up an account - and the rest of your data is already there!…

I'm not a lawyer but my understanding in the United States is that a person in a public place has no legal expectation of privacy. If you are in public and someone takes a picture in which your face is visible and then upload that to Facebook, I wouldn't expect that to be a violation of my privacy nor would I expect Facebook to allow an unknown person to request the picture's removal.

I'm not form the US and also not a lawyer but I've heard that this goes even to the point that someone using a public communications network has no legal expectation of privacy in the US.

The lack of data-privacy regulations is likely rooted in the fact that the US never had an experience like the EU with the Nazis who literally killed on behalf of data.

Sure, the US famously "kills people based on matadata"¹, but not US people so most US people don't care.

But I've also heard that some people in the US start to recognize the value of data-privacy regulations. That's positive to note.

¹ https://en.wikiquote.org/wiki/Michael_Hayden_(general)

Re: Facebook is researching AI systems that see, hear, remember everything you do

#130

Post by lobocinza »

istorical wrote: post #48
hetspookjee wrote: post #37

Sometimes I dream of a Facebook machine; e.g. the eerily creepy recommender for your next best action, but than with an objective function not to optimise the recommenders revenue (by being as effective ad seller as possible) but rather changing the objective function to optimise for health, wellbeing and happiness. Now surely it is far from trivial to define these, but atleast you can make a start somewhere: - You c…

I find it absolutely painful that our recommender ML based content discovery streams like YouTube and Tiktok and Instagram don't have multiple 'hats' you can put on and take off, in the sense of saying 'hey TikTok, I'm feeling frisky, feel free to send me those thirst trap videos now', but then 30 minutes later you can say 'OK TikTok, I know I watched a bunch of those dancing videos, but can you send me the education…

You can use YouTube anonymously and also clear your history (which I frequently do). In my experience the algorithms are good when they know just a little about you and make good recommendations once in a while. Once you watch too much of something it become a rabbit hole of the same stuff but more dumb.
Post reply on HN