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

https://www.theverge.com/2021/10/14/22725894/facebook-augmented-reality-ar-glasses-ai-systems-ego4d-research

by coldcode • 5 years ago

189 131 5 years ago

h2odragon• 5 years ago

Apparently whoever paid for the most recent "Facebook is Bad" media hate package, only bought the 3 day ticket. Cheap bastards.

lucasverra• 5 years ago

Who would benefit from that service? Maybe PR services companies? "See what is happening to Fb, we can help you avoid that Mr other dodgy Tech"

Who else?

blackoil2• 5 years ago

Traditional Media. Politicians. Competing Social Networks.

Also other tech. companies as FB is sucking up all media/political/people's energy and the rest can cruise current wave unharmed.

knownjorbist2• 5 years ago

Facebook is consciously aware of the fact that they're optimizing a manipulation engine that has outsized net-negative effects on societies worldwide.

arpa• 5 years ago

No, thank you. But I suspect people would still accept and use that tech just like they have accepted always-online/always-listening digital assistants.

Just imagine the possibilities of large-scale manipulations, tho!

fctorial• 5 years ago

Do you use history in your shell?

CGamesPlay• 5 years ago

I do, but I don't upload it to any other parties.

danShumway• 5 years ago

What exactly is the comparison here?

Remembering my shell history for commands I run locally on a single computer, stored in a format I can control/edit, that is never transmitted to other people or used for advertising, that is fed into a simplistic history algorithm I have complete control over, and that I can toggle on and off at will even for individual commands -- that's the same as a 3rd party analyzing on remote servers everything I look at while I'm walking around the real world?

0xdeadb00f• 5 years ago

Only for the current session. When I close the terminal it is all gone

seadan83• 5 years ago

Yes, and you can lead a command with a space for it to not be remembered in history. You can also delete it, and nobody else can access it without privileged access to your computet

sys_64738• 5 years ago

I hope it remembers why I deactivated my FB account.

hungryforcodes2• 5 years ago

I wish you so many upvotes for this comment.

throwinawaysoon• 5 years ago

upvotes: thoughts and prayers of the internet

mlok• 5 years ago

When you deactivate your FB account they keep all your datas. I hope you know that. Deleting your account should delete your datas, though. (But I don't 100% trust they do. And maybe some national agencies keep a copy)

dataviz1000• 5 years ago

So you are saying that you hope it remembers why you went to https://www.facebook.com/deactivate where you deactivated your account without deleting your profile?

Gys• 5 years ago

You will be replaced internally by the same AI impersonating you. To keep the ads running.

jmfldn2• 5 years ago

When I read articles like this I experience a mini existential crisis that we could be heading to some very dark places technologically.

I don't need to go into the reasons why it's bad. For anyone who's aware of the last 15 years of social media critique or novels like Brave New World, it's obvious that this would be a dystopia multiplier. Ad tech on steroids. If you think the manipulation is bad now just wait until this data is fed into the ML models.

If you must have everything you see, do and say recorded then please, for the love of god, let's use non-profit, open source / free software platforms where we own our own data.

sneak4• 5 years ago

In one generation in the US we went as a whole society from most private p2p conversations not being logged and monitored and recorded by the government, to most private p2p conversations not only being logged and monitored and recorded, but available to the state at any time without probable cause or a warrant (FISA Amendments Act).

iMessage, WhatsApp, Gmail, Facebook Messenger, Instagram: all of these are stored on the server effectively unencrypted (in the case of iMessage and WhatsApp, via the chat backups, a backdoor to the e2e encryption they promise). The state has access to all of them at any time. To top it off the carriers and a million different apps are constantly logging your location history for all time, for every member of society.

If you think that isn't one of the most useful and powerful databases in the history of mankind, I question your imagination.

The massive damage, which may just be existential, is not apparent to the USA just yet, but it will be eventually.

The balance of power has shifted even more dramatically than just about ever before in history, in any country in history. Alarm bells should be going off left and right but nobody seems to care and it's just business as usual.

jmfldn2• 5 years ago

Agree. We desperately need something like Tim Berners Lee's Solid. A way for us to take control from the centralised data silos owned by big tech. I know its almost mission impossible to turn the tide, but if the alternative is giving in then engineers worried about this must fight for alternatives, however grim the odds.

As for this specific issue, of course there are the usual massive issues around data privacy. The major issue I was alluding to here is ad tech AI processes being fed an unimaginably rich source of intimate data. It's a horrifying prospect.

01100011• 5 years ago

most private p2p conversations not being logged and monitored

FWIW, I think telephone company billing records have been doing this for nearly a century. Sure, it wasn't the official government policy, but the data was recorded. ECHELON has been around quite a long time as well, and that was indeed government monitoring of large amounts of international communications.

I'm not trying to normalize it, just adding some context.

LuisMondragon• 5 years ago

AFAIK the chat backup is optional. I've never stored a backup.

ianlevesque• 5 years ago

WhatsApp just fixed that by encrypting its backups too. https://faq.whatsapp.com/general/chats/how-to-turn-on-and-tu...

decasteve• 5 years ago

Social media have become Bradbury’s parlour walls.

908087• 5 years ago

As a society, we've been sleepwalking into a very dark place technologically for well over a decade now, and those of us who have tried to point that out have been routinely shouted down and called luddites for it.

I fear we're far past the point of no return, and there is no escape from the disaster we've created even for those of us who have avoided participating ourselves.

radmuzom• 5 years ago

In addition to being a "dystopia multiplier", I predict this is going to cause a lot of social problems going forward. This is what the Black Mirror episode "The Entire History of You" warned us about.

still_grokking10• 5 years ago

I'm already looking forward to brain-internet interfaces.

Does FB invest also in those?

Just imagine, all the possibilities!

[Do I need to mark this as sarcasm?]

mgreb• 5 years ago

Wait, are such devices even legal? Looks like walking spy-cam to me. I want to know if someone records video/takes picture of me. With this glasses you can`t tell if you are being filmed.

blackoil2• 5 years ago

Is it illegal to shoot videos in public place without permission of all present ? Private places may have their own rules. Also this is R&D, long way before such questions matter.

jjulius2• 5 years ago

Nah, these questions matter now.

de_keyboard• 5 years ago

In most western countries it is legal. I'm not sure it should be, given recent technological advances.

picardythird3• 5 years ago

Don't you think answering these questions should guide the R&D?

KaiserPro2• 5 years ago

Yes, there is a "right to panorama" in most countries. However this dataset has been recorded with the _express permission_ of all the people involved. Which is different to the imagenet, where they just crawled a whole bunch of images and called it a day.

Second, you need to take the "Argh its facebook, boo hiss" hat off. Then apply some critical thinking. an AR world that is connected to the physical _requires_ this kind of thing. Your device need to anticipate what you are going to do so that it can work out the probability of your need and act on it. Like Jeeves, but less able, and more annoying. As its ML, it need a massive dataset to train on. this is <0.5% of that dataset.

Depending on how things are done, if facebook are first to market with a usable AR system, they will be forced to have anonymisation built in (as in remove faces at the sensor level, unless you have permission to remember). Apple will have a showy "pixelation" layer, that is mostly ineffectual, but will PR it out to make them seem like they've cured cancer. Google, if they ever manage to get back into AR will just make things cheap and let the shitty android marketplace spy on what ever they like.

You will also have to remember that the power budget of these glasses is absolutely fucking tiny. All day screen, SLAM, AI and possibly music will need to all fit in to ~2-5watt hours. This means that virtually everything will need to be on device. (dropping to the cloud eats a boat load of power.)

Now that's not to say that 100% accurate [if thats even possible] "diarisation" won't rip society apart.

There are lots of questions that need to be answered, the problem is, tech journalists are ill equipped to ask them. Most of the teams designing the glasses are well meaning, but their targets and life experience has not equipped them to do a good job at ethics.

kwertyoowiyop• 5 years ago

All I need is to see people’s names right above their heads.

jstx1• 5 years ago

It should only cost you half of your remaining life span.

jensensbutton• 5 years ago

Anyone mad at the universities that actually collected this data? Or just Facebook?

de6u99er• 5 years ago

I am mad at both. Universities should not participate in such unethical activities!

Just as a reminder Cambridge Analytica also started at a University. And many politicians are seek advice from academics who are directly or indirectly incentivized by Facebook. Through grants and/or through access to data others don't have access to.

fithisux• 5 years ago

... and don't forgive, unless you buy the "premium" version.

ToddWBurgess• 5 years ago

And this is why you need ad blockers and privacy protection plugins for your web browser. If you can't stop Facebook snooping on you at least make it hard for them. As my grandfather taught me, "if you can't win then make it hard for the other guy to win."

kelseyfrog2• 5 years ago

I don't like this pattern in calls to action. The pattern starts with an implicit hopelessness of affecting any external change and results in all responsibility being assumed by the individual.

This is an antipattern of change. It does not need to be perpetuated because it reproduces the permanence of the constructs we want to dismantle. Even more importantly, these entities are fully aware of this and craftily guide the masses toward individual responsibility as a default, knowing that it won't affect real change. The antidote to this is critical collective action which realizes the impossibly of change as itself an impossible stance to take.

pessimizer2• 5 years ago

Change isn't caused by strength of belief; it's magic that only works based on how hard you believe in it.

Collective action doesn't do itself, and the words "collective action" don't constitute a plan. Individual evasions at least make things more expensive to do.

Also, individual action doesn't crowd out collective action. Not protecting yourself doesn't create collective action. A concrete plan with concrete individual actions to take creates collective action. Individual sacrifice to create institutions creates collective action. IMO you should be ready to tell people where to show up before you tell them not to help themselves.

dylan604• 5 years ago

It just seems like the people at FB have watched Black Mirror and decided they want to do all the things they saw.

shimonabi• 5 years ago

That "social credit" episode was really creepy.

kwere• 5 years ago

or they became DARPA lifelog 2.0

blunte• 5 years ago

Regardless of who provides this, and even irrespective of marketing concerns, there most likely will be data leaks/theft. Imaging having audio+video recordings of what you see and what you say in the hands of the wrong people. And "the wrong people" could be a wide array of different parties, from banks to police to political enemies to nosy neighbors and beyond.

Granted, many of us already collect and give away significant amounts of very personal data, some of which regularly gets leaked or stolen; but first person video and audio recordings?... scary.

And then with Facebook... I trust Zuckerberg with my life data as much as I trust his choice in hairstylists.

fergie• 5 years ago

There is no way that this will ever be legal in the EU.

hetspookjee2• 5 years ago

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 could factor in the cost for environment per each recommendation and have that drive down the score of a recommendation: I think the "fast-fashion" idiotic anti-environmental trends would die out fast than.

- You could factor in, based on some overarching theme of ethos, that each person needs to eat healthy and recommend the right healhty food at the right time; instead of recommending the next best McDonalds burger to an already obese person.

- Take into consideration, based on studies, what actual heatlhy behaviour looks like, and try to fit that to a person - with respect to their actual persona and state of health (which are obviously known already given that each bit of you is pretty much on the street already)

I could go on, but in truth, a personal recommender that would know me better than I even would, and would optimise towards happiness and greatness, instead of # shit sold like it currently is, would already be a tremendous improvement. Surely, this idea is also far from original, but I hope that we someday get to such a thing. Atleast a thing that stops recommending environmentally destructive behaviours like buying so much stuff I don't really need.

SamoyedFurFluff• 5 years ago

Unfortunately I think we live in a society where the people who could use this the most are the people least able to afford the costs of such a service.

rumblerock• 5 years ago

I'm certainly wish this were possible, and people who left would probably go back to the platform if it reformed itself in this direction. But as the years go on all I see are more and more catastrophic impacts from short term thinking and near term profit generation, whether it's Facebook, climate, VC-funded startups or the current supply chain crisis.

Perhaps the public consciousness around the unhealthy design of existing social media is approaching a point where a new Facebook-like social network could actually achieve a critical mass of users. On the order of years I think it'll happen (like rates of smoking cigarettes), but how long that will be who knows. I can only hope it comes in the next few years as I fear the destructive ripple effects through culture and society are only getting worse.

istorical• 5 years ago

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 educational stuff now'. Ditto IG, YouTube etc. Like 'hey YouTube, engage education mode', 'hey YouTube, engage self-improvement mode', 'hey YouTube, I just wanna relax and laugh'.

I know that there are categories, even auto-generated categories on YouTube, but they aren't that great or all encompassing. My biggest critique of present personalized ML content streams is that there's no two-way communication. I can't tell IG that yes I wanna see big booties right now, but no I don't want to see them five hours from now. Or I want to see aquaponics right now since I'm on a learning binge, but later tonight I don't need more technology and science, I just need to laugh and unwind.

Our actions, our content consumption, is not allowed to be one and done. Every move you make is seen as a signal to the algorithm that you want more of that all the time. The closest thing we have is incognito mode, or maintaining multiple accounts for different purposes. But why can't you look at an ex on instagram one time without them showing you that person everytime you open the app for 5 years. It's really toxic and unsustainable that all of your actions are seen as 'yes, please more of this all the time'.

jorpal• 5 years ago

Anyone remember StumbleUpon? You could easily edit your interests and get different types of content. It was the pinnacle.

clairity• 5 years ago

"...in the sense of saying 'hey TikTok, I'm feeling frisky, feel free to send me those thirst trap videos now'..."

i once pitched this type of user experience (for the general case, not tiktok, which didn't exist at the time) at a hackathon competition, and it was popular enough then to win us some free stuff. the demo we hacked together was very rudimentary, but had we gotten a little further along with a proof of concept before falling apart, who knows what might have happened...

the biggest challenge was actually licensing costs and rights management, not the recommendation engine (though that was challenging too).

lobocinza• 5 years ago

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.

mertd• 5 years ago

changing the objective function to optimise for health, wellbeing and happiness.

Whose? Individual or group wellbeing? Optimizing for one can lead to very suboptimal outcomes for the other.

tomjen32• 5 years ago

This feels like the argument about weather the self-driving car should hit two grandmas or the guy in the car. The discussion is not productive, because the answer that saves the most people is to get self-driving cars on the streets as soon as possible.

As for the answer to your question, just go for best wellbeing for the individual. Happier, healthier, less obese, fitter, non-smoking people are also a net benefit to society.

Assuming the technology works at all.

dado3212• 5 years ago

FB has gotten roasted for doing emotional manipulation in the past: https://www.forbes.com/sites/kashmirhill/2014/06/28/facebook.... Not confident that the company trying to specifically target emotions (even ostensibly happiness) would be taken that well.

Frost1x• 5 years ago

I'm not sure it would matter. The general population that compose Facebooks revenue stream don't seem to care. We keep hinting at these ethical conundrums that invisibly regulate massive businesses when they simply don't. It's not just tech, it's big businesses in general. The ethics only matter if a large enough population of the consumer and product base both care and act. Only demand matters and if demand doesn't care or you can manipulate the demand not to care, you can get away with about anything that isn't explicitly illegal.

I suspected Facebook does the work you referenced but I had no idea they actually did and had any flack about it. I'd like to think I'm a fairly tech savvy and an informed consumer (especially around issues like this), maybe I'm not, but if I am, what hope do you have of the general population caring when they aren't even aware as a first hurdle before the other massive hurdles of getting them to care enough to act.

The fact is that this type of work isn't regulated. Funding agencies regulate human subject work for your typical public funded research work and require a lot of informed consent. The leverage they use is that should you violate these terms, you may have current and future funding pulled, may be black listed across multiple agencies, and may find it difficult to pursue any career in research in the future involving human subjects. There are lots of disincentives not do this on the premise of manipulating your resources.

Meanwhile, entities like Facebook have piles of such resources and have a vested interest in all sorts of this questionable work. Pretty much nothing prevents them from doing it since they're self funding. Unless you can remove their self fundability, it will continue. This goes back to problem of needing consumers to vote with their wallet in an effective manner.

The other option is to create the explicit policy protections in law to regulate these activities but as a society, capital has convinced everyone that all regulation is evil and will only hinder consumer/citizen progress. You need not only policy but real enforcement teeth as well that create disincentives that are catastrophic enough not to tempt risk for the potential gains.

hetspookjee2• 5 years ago

I do not believe Facebook as a company would be possible in generating such a thing. And if they could I can only be cynical about it. I rather put my faith in the open source community where people tinker on their own algorithms, and provide their own data. I am sure the first initiatives are already on their way regarding personal recommenders. But I don’t know any. If anyone does, please recommend :)

pesenti12• 5 years ago

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...

Hnrobert423• 5 years ago

What are you all doing to allow those of us that don’t want to be a part of your utopia to opt out?

pesenti12• 5 years ago

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.

nbzso• 5 years ago

I support free and open web. We are in defining point in time in which a possible tech dystopia is upon us. No paycheck or "advancements" will remove the dangers and damages done by greed and machinations of few to Big to fail companies. No dataset or "business" existence is rationalization for creating a system of data-hoaridng and data exploitation.

dannykwells• 5 years ago

This comment needs a serious disclaimer that you are a VP of AI at Facebook. Its implied in the above, but such serious COI requires acknowledgement.

amelius• 5 years ago

Glad for you that you collected such a nice dataset.

Wouldn't it be nice if selected external researchers too could collect data on the Facebook platform with the explicit consent of users?

pesenti12• 5 years ago

Here is a link to all the tools we provide: https://research.fb.com/data/

knownjorbist2• 5 years ago

Maximizing a manipulation engine is not what we as a species need to be investing any time or effort into.

dredmorbius3• 5 years ago

What is your price?

Who knows it with greater accuracy, you, or FB?

stereoradonc• 5 years ago

I support Facebook shutting down this program.

goldenkey• 5 years ago

It's so strange how everytime I get a Facebook link from someone, it requires me to login to read more than 1/4 of it ( and it's cut off just at the right place to entice ) and then before I know it, I'm looking at a bunch of silly unread notifications and terribleness in my feed. It's strange, like a realizing I was hypnotized for a second. I immediately log out and close the window as I snap out of it. Facebook is truly...a cancer. Do not give me the whole song and dance about how their targeted ads help small businesses. 99% of the stuff I've seen on FB, business-wise, is ads for shady products, scams or schemes.

They took a truly golden opportunity to connect people and ravaged it, then turned to acquiring every possible competitor they could, as their main product turned into hot garbage.

Even AOL garners more of my respect than FB, how is this even possible?! FB needs to die like AOL died. It'll only take the boomer generation to die out. Unfortunately, Instagram is still a decent property, probably because they didn't let Zuckerfuck fuck it up. Expect Instagram to become cancer too once FB is no longer the main profit center.

achenet• 5 years ago

This reminds me of the Ted Chiang story The Truth of Fact, the Truth of Feeling.

sabujp• 5 years ago

if you use fb, unfollow everything and everyone and then either selectively follow what you like or don't ever follow anyone again : https://gist.github.com/renestalder/c5b77635bfbec8f94d28#gis... , it will greatly reduce your time on fb.

TeeMassive• 5 years ago

Episodic memory: What happened when (e.g., “Where did I leave my keys?”)?
Forecasting: What am I likely to do next (e.g., “Wait, you’ve already added salt to this recipe”)?
Hand and object manipulation: What am I doing (e.g., “Teach me how to play the drums”)?
Audio-visual diarization: Who said what when (e.g., “What was the main topic during class?”)?
Social interaction: Who is interacting with whom (e.g., “Help me better hear the person talking to me at this noisy restaurant”)?

Now consider the fact that Facebook is actively suppressing some subjects from popping up in their users' feed. If users becomes reliant on this AI to initiate, guide and execute most day to day tasks then Facebook will literally have the power to make certain aspects of their users' life disappear.

For example, you have a contact who criticizes Facebook or their business partners too much? Don't remind them of their birthday.

moogly• 5 years ago

WW3 might be against Facebook.

zwaps3• 5 years ago

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.

reilly3000• 5 years ago

I’ve been hanging around HN for quite nearly a decade. This is the first comment I’ve felt compelled to mark as Favorite. Precision perception and behavior modification at scale is a dream of autocrats and marketers alike. The fact that it is private/anonymous is entirely irrelevant.

peakaboo• 5 years ago

I don't think humans are smart enough to avoid this fate. The future will be exactly like the dystopian movies we are watching because that's what the megacorps want.

It's pretty simple to see, but a lot of people don't want to see it.

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

Related Stories