Nowadays, a professional engineer (a phrase with a precise legal meaning) can be held criminally and civilly liable for negligence even when they are "just following irders." Unfortunately, we're at the very early wild wild west stages of the industry, before some big exposé or calamity completely shakes up the industry (i.e. Sinclair's The Jungle did for food, Carson's Silent Spring for pesticide use, or the 1906 SF earthquake for building codes).
I think of these algorithms as being algorithms created and vetted out by algorithms, therefore, its negligent to not understand what they are doing. Since its difficult to know what they are doing, then the risk to use them in this way should be viewed as too great
The only way to ensure nobody is killed is to add "0 people killed" to the target function of the black box.
peopleKilled() < εIt would make a good novel, but it would be too confusing and depressing, like the Infernal Desire Machines of Dr. Hoffman, but more frustrating and confusing.
Facebook is trying to get people to be miserable because then they are using facebook more, and its engineered in such a way that it makes us addicted to facebook and want to use it more anyway.
Is this enough reason for our society to revolt against facebook? How do we actually make it go away when everyone is addicted to it. Does anyone have any practical tactical strategies to get rid of it?
It's a bit like a parasite - it needs to control and extract usefulness from the host but not let the host die quickly because well it won't get the resources anymore and it would die with it.
> Does facebook have enough power that it can chose to kill off certain people? A
I never signed up for Facebook to start with, but I got a glimpse of this power with Netflix. The movie recommendations were so good that I end up liking them if took its advice. Where normally I wouldn't have watched them based on rating, synopsis, or title. I had realized some algorithm at Netflix seems to know me better than I knew myself. Then thought what if it suddenly decided to use that against me, because it somehow would be more profitable for it.
It's like time spent watching TV. I enjoy it; sometimes it provides value; it's tendency is to take over and create an unhealthy life for me.
Anyways even if this friend posts about visiting on facebook, this is not guaranteed you will be shown this post, as demonstrated by the actual HN submission you're commenting about.
And yes, there was no facebook 15 years ago. That is great, 15 years ago everyone new to SMS or email you. Now, everyone expects you to be on facebook. See the difference?
I'm lucky in the sense that I made a FB account, tried it out for two weeks, realized it wasn't for me and then cancelled so I never got hooked. But I can see that once you are in it takes a bit more effort to get out. But being 'on Facebook' should not be a prerequisite for normal social interaction.
Why not? Collectively it's proven to be the easiest way for groups of loosely and moderately connected people to communicate and coordinate socially, so it's what virtually everyone I know uses. We all use Facebook for events and for Facebook Messenger. The set of friends who live within train/car distance who I communicate with via something other than Facebook Messenger numbers around...four? Out of dozens? And it's not because We Don't Use Facebook, it's because we share a Slack and had an IRC channel before that.
By not using it, somebody my age and in my neck of the woods relies on making people who they ostensibly call friends treat them as a special case just to organize beers on a Friday. And it does make that person a special case: why would people use group-SMS anymore when Facebook chats are right there and you don't need to track down phone numbers to include everyone (and deal with sharded SMS threads if people are added, etc.)? The expectation that people special-case your edge-case life decisions has to be balanced against not being a pain in the ass for people you claim are your friends.
So, yes, in many quarters Facebook is a prerequisite for normal social interaction (unless you're That Guy, and it's good life advice to never be That Guy). And, TBH? It works fine. It's nothing special. It works fine. I very rarely see a news feed (I use Social Fixer) so I have no real opinion on the way it does likes/shares/etc.; I did tag some people as "close friends" so when they post it does reliably surface, but I don't remember the last time I even scrolled down the news feed and its existence has no impact on the other features of the platform.
It is just not that big a deal. It's certainly not worth the airs put on--not from you, but the I-don't-even-own-a-TV levels of smarm from people who don't use Facebook makes my eyes roll out of my head and bounce down the stairs--by people who don't use it.
>Why not? Collectively it's proven to be the easiest way for groups of loosely and moderately connected people to communicate and coordinate socially
I'd be happy to have a source for this outlandish claim.
As to answer your question, there are many many reason not to let a private transnational company take control of your social interactions. Easy one: what happens to your social interactions when facebook disappears as every other so called social network before it did (remember friendsted, myspace, etc.) Another easy one: what happens when your account gets locked and you get banned from facebook ?
Harder ones are not so obvious for people who do not stay informed such as the deep impact on shaping people opinions and thoughts, the disappearance of empathy, intelligence dropping, destruction of internet, support of foreign dictatorship, massive tax fraud and tax evasion, emotional manipulation, organized abuse, sextorsion, and a lot more.
My entire friend group and extended friend group in the Boston area. Somewhere between ten and twenty thousand people within two hops of me and in the local area, geometric expansion's weird like that. If I want to talk to any of them, we will end up using Facebook. If a friend intros two people face-to-face and they want to talk later, we use Facebook. If I meet a girl at a bar, or if I meet somebody who might be cool to work with on a project, and we want to hang out later, we use Facebook. Suggesting SMS is usually tolerated but makes you a little weird and makes you harder to get ahold of and you're likely to be ignored and left out later; suggesting email (outside of an explicitly business context) makes you a dinosaur and makes it very likely that you are ignored.
I cannot remember one time in the last five years that somebody suggested SMS as a contact method to me. I have phone numbers for folks I knew before Facebook was a thing, sure. And phone numbers are a good-enough "front door" for stuff like online dates. But actual interaction? Overwhelmingly, overwhelmingly via Facebook Messenger.
Nobody's saying you have to use it. But, in my experience, making it harder for people to contact you means people don't. The fumfuh about "well my real friends" ignores that there are both social and transactional benefits to not being sand in the gears, not being a pain in the ass, for people who are not your real friends. Being sorta friends is still both fun and useful.
> As to answer your question, there are many many reason not to let a private transnational company take control of your social interactions.
The assertion I am making alongside that rhetorical statement--which, TBH, I'm not quite sure why you thought was not rhetorical enough to need a soapbox stand--which is missed in the less friended corners of HN, is that all of this matters a lot less than keeping in touch with people.
Datamining--whatever. Oh, no, Facebook knows I go out for a drink on Fridays. Oh, no, the like button on pages (which don't actually phone home, because I have a hosts-based ad-blocker on my phone and my browser) shows that I read the Times. The tradeoff of that whatevermining versus not having a functioning social life in the year 2017 is immense and the types who don't-own-a-TV about Facebook display a distinct lack of understanding that maybe that trade-off is okay for some people and that it does not necessarily come from ignorance but is weighed and calculated--because having a social life is really nice and not having a social life really sucks.
Facebook is often annoying and is kind of shitty, sure. It is also the only game in town. In plenty of social circles, you play or you don't participate. The lack of understanding and empathy out of the "oh, just don't use it" camp is not shocking--tech people choosing not to understand things they might not agree with and choosing not to employ human empathy is not and never will be shocking--but it remains profoundly annoying.
It is really, really good at that stuff, and you can--you really can!--pick and choose what you want to use and what you care about.
It's akin to "I've used this email address for years and decided to stop checking it and use another one but did not tell anyone about it. Now I'm missing messages people send me to my old address I do not check anymore".
[0] A decent review with code, a Noble laureate, and a top notch Computational Economist: https://lectures.quantecon.org/jl/rational_expectations.html
I have no interest at all in a solution that can’t be explained. That’s not a solution, that’s a remedy.
I don’t have a lot of time for humans that know things but can’t articulate them. Why would I make time for a computer that does the same? It doesn’t even earn the courtesy of give a human.
I was drawn into this field by the idea of understanding things and solving problems with them. If AI can’t “show it’s work” homework style, then it might find me a shorter route to the mall but it hasn’t solved anything.
Not all things that humans learn are easily articulated or explained, not because those humans aren't articulate, but because they learned actions that are more easily intuited.
Also, on this topic, explainability of machine learning models is a hot field: https://www.youtube.com/watch?v=KP7-JtFMLo4 https://arxiv.org/abs/1602.04938 https://github.com/marcotcr/lime
I agree, but far from being an unique or surprising property of Facebook and the way it (as a company) works, I'd say it's almost in the nature of any system built by a large enough group of people that "nobody there really understands how it works".
Physical products have reasonably comprehensible user experiences. That's even true for much software. Microsoft Word is a beast, but the user experience is tethered by the external reality of what people do with it. Memo writers and novel writers and PhD dissertation writers all have different needs, but it's reasonably clear what their goals are and what successful outcomes look like. There are approximately equivalent tools that can be swapped in.
Facebook, on the other hand, is creating an ongoing experience. My needs are jumbled together with that of my friends, their friends, strangers, news organizations, advertisers, companies, non-profits, celebrities, would-be celebrities, and Facebook itself. And each experience is deeply individualized, adding yet more complexity.
It still probably had some intellectual coherence at a much smaller size. But they have so many people working, and so many of them beavering away at optimizing semi-conflicting metrics based on so much data that I don't believe it makes sense to anybody there. Much less sense than Word makes to its creators, for sure.
Sure, there are ways to ask some ML systems why they made a decision after the fact, and they can elaborate which variables had the most effect. But before the algorithm gets the training data, you don't know what it will decide - nobody does!
It picks actions that maximise a value it was told to maximise. It does this based on "well, when I encountered a situation like this other times, I did these 5 things, and this one thing maximised that value I'm told to maximise in most of those cases, so I'll do that.
At the beginning it will try random things. After several iterations, it will find things that "stick" i.e. maximise the value. There are sophisticated techniques to get unstuck from local maximums and minimums, too, in order to find even bigger maximums.
When you have enough data, enough users, enough actions in a system, you can very quickly try a lot of things and learn a lot of stuff. But what those stuff would be, no human can predict.
The main problem is that its unethical to train a machine learning system with a value function that doesn't take into account basic ethical behaviour. There are 3 challenges here
The first challenge is technical, and its the easiest one - how to come up with a value function that encodes additional measures of human ethics. Its easy as in, you take a bunch of ethicists, give them various situations and ask them to rate actions as ethical/unethical. Then once you have that, you try to put it into the value function, so that the system can learn some decency. All done. (If only it were that easy!)
The second challenge is a business one, which is, how far are you willing to reduce your value maximisation to be ethical, and what to do if your competitor doesn't do that. One way to solve that is to have regulations for ethical behaviour of machine learning systems. The system could be held responsible for unethical actions. If such actions are reported by people, investigated by experts and found true in court, the company is held liable. Public pressure and exposure seems to help too. Perhaps we could make machine learning systems that detect unethical behaviour and call it the ML police. Citizens could agree to install the ML police add-on to help monitor and aggregate behaviour of online ML systems. (If these suggestions look silly, its because they are)
The third challenge is philosophical. Until now, philosophers were content with "there is no right answer, but there have been many thoughts on what exactly is ethical". They better get their act together, because we'll need them to come up with a definite, quantifiable answer real soon. On the more optimistic side, lets hope that any generally agreed upon ethical system is better than no system at all.
i propose that all legal actions are ethical, and all illegal actions unethical. There shouldn't be actions that are considered unethical, but legal (and, for completeness's sake, no ethical action is illegal either).
Then, society will need to continue to codify their ethical framework as law.
Laws are based on ethics, but not everything that is unethical is also unlawful.
It's usually legal to lie, break promises, exaggerate, and other things that I would consider unethical. I also submit that the should remain legal and unethical. Lying to my parents about some detail of my personal life I don't wish to share is unethical, but shouldn't be illegal.