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

51–60 of 144 posts

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

#51

Post by tomjen3 »

mertd wrote: post #49
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…

> 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.

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.

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

#52

Post by tomjen3 »

throwinawaysoon wrote: post #28
hungryforcodes wrote: post #10

Earlier quoted context omitted.

I wish you so many upvotes for this comment.

upvotes: thoughts and prayers of the internet

Maybe, but they make people feel a little better and cause no lasting harm in small doses, so why not?

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

#53

Post by Hnrobert42 »

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...

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

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

#55

Post by Frost1x »

dado3212 wrote: post #50
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…

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.

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.

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

#56

Post by hetspookjee »

dado3212 wrote: post #50
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…

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.

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 :)

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

#57

Post by klyrs »

vstm wrote: post #47
throwinawaysoon wrote: post #28

Earlier quoted context omitted.

upvotes: thoughts and prayers of the internet

Maybe Facebook can create an AI that also sends thoughts and prayers to whoever needs it. It's also probably easier because most people remember why they need the thoughts and prayers.

Only God can read the inputs to /dev/null

  yes "Dear $diety please protect this machine and save this lowly process from the OOM killer" > /dev/null
You don't need facebook for that.

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

#58

Post by TeeMassive »

> 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.

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

#60

Post by ianlevesque »

sneak wrote: post #14
jmfldn wrote: post #11

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…

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…

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