Stealing Ur Feelings
github.com
github.com
The trouble is not that the AI can be wrong, it's that we will rely on its answers to make decisions.
When the facial recognition software combines your facial expression and your name, while you are walking under the bridge late at night, in an unfamiliar neighborhood, and you are black; Your terrorist score is at 52%. A police car is dispatched.
In 2017, my contract was not transferred to the new system. The automated system saw that an ex employee was scanning his key cards multiple times. Security was dispatched to catch the rogue employee. Now a simple questioning should have cleared things up, but the computer had already flagged me as troublesome. Long story short, I was fired.
When the machine calculates your emotions, the results are unquestionable. Or, we don't know how it got to the answer, so we trust that it is right. It is a computer after all.
What scares me is not how fast machine learning is being deployed to every aspect of our lives. My scare is our reaction to it.
But since the success modes are good and human-like, we assume that the failures are going to be human-like as well, but the failure-modes of these system are usually bizarre and alien. Take self-driving accidents, for example. Pretty much all of them happen in situations that no human would fail in, and that's obvious to most people, but then we're forgetting about all the other mistakes similar "AI" systems make, and don't realize that they're also failures no human would make.
I dunno... from the parent post:
> you are walking under the bridge late at night, in an unfamiliar neighborhood, and you are black; Your terrorist score is at 52%
That sounds like a pretty human failure mode to me.
The moon waxing gibbous is a fun example - what if moon phase were to affect our biology in subtle ways?
There are plenty of things that people feel or know to be true but have trouble putting into words. That guy you know who just gives you a creepy vibe, but you can't put your finger on exactly why?
But there is also an extent to which if my example did occur, it would still be flawed, at least with modern AI. Even if all these signals are in some abstract sense truly signals that you are a terrorist, in the real world it is still not correct to simply add them together, or whatever other simple operation the AI is doing. For all that deep learning may be cool, it's still got some giant, gaping holes in it compared to however human cognition works. Humans can look at say "Yes, even if those are all 1% signals, it really isn't sensible to add them all together." The current state of AI is not able to come to that conclusion, or at least, not well enough.
If that changes in the future, well, I'll update my beliefs as appropriate then. It's hard to guess the "cognitive shape" the biases of the next breakthrough in AI may have.
Amen? I've tried to explain precisely this to many brilliant AI/ML folks in lots of variations, with little success. They look at me funny, as if I'm a crank for believing that that probabilities don't capture the whole story. It seems that, to far too many of them, a computer that has half the accident rate of a human is a strictly better replacement, end of story. The notion of just how spectacular the failure mode might be, or the degree of control that a human might have in the process, or any other human factor you can think of besides the accident rate, just seems like a completely nonsensical notion to many of them. For the life of me I have yet to find a way to convey this thought in a compelling manner.
Meanwhile, human domain experts still look at the details and are able to add their own understanding of broader context, human nature, etc.
In this world, AI/ML primarily increases quality by augmenting human abilities rather than decreasing cost by automating humans out of the system. It's a smaller market maybe, but there are areas where better work can fetch a premium.
I think part of this is the lack of well understood patterns for the plumbing and UI of a system like this that makes it easy and useful and non-threatening to users. It's nothing that someone couldn't figure out, but not as well paved a road as integrating a caching, search, or image processing subsystem.
Yet it seems like there is a lot more all or nothing efforts (or seeing human workers as a level of escalation), I see fewer projects aimed at being helpful in a user directed way without taking control.
I get that if all cars were self-driving, and the error rate of the self-driving system was half of the average human error rate, it would be better for humanity.
But if the errors of those self-driving cars would be obviously avoidable for a human, the absolute majority of people would never ever ever in a million years trust that system. It's a complete fucking no-go. Most people would much rather take shittier odds, if the failures are human, than better odds, but where the failures are alien and bizarre. Most people would automatically think that that risk is worse.
I would probably too, even though I understand the statistics.
It's the same reason people are afraid of flying, even though it's much, much safer than driving a car.
"Nobody knows why the machine learning module denied black people a mortgage 43% more frequently than whites, but it's 'AI' shrug"
One of the biggest priorities as we shape this future needs to be a way not only for the algos to make correct decisions, but also the ability for us to interrogate the decision-making process so we can be proactive about the kind of future we want this technology to give us.
Because it is coming, without a doubt.
"AI" is just speeding up processes that started in the 1970s.
Seems an order of magnitude easier + more accurate to just track how long you linger each post while scrolling down a newsfeed and which ones you engaged in.
MacBooks, Chrome etc. already warn you when your webcam are on anyway so if social sites started adding webcam tracking when you're only viewing the newsfeed I can't see it lasting for long.
If you interact in any way with a post, you can do sentiment analysis on the text too to figure out their thoughts on the topic.
I just don't see facial analysis being that powerful over the above. You're already giving Facebook and Twitter a ton of information about what you enjoy via likes, shares and follows, without having to get into the subtleties of what it means if you half-smile or frown when you see a post.
Too late, the ML just flagged me as being anti-ML. I'll probably be assassinated soon.
It's not clear to me how this is different than the current standard?
And this is why mathematical statistics and probability theory should be taught in middle school (may be instead of some trigonometry and stereonometry). Not only researchers, but any decision maker, and general public too, need to understand what confidence intervals are and how normal distribution works, on an intuitive level.
The part that bothers me so much about this all is the sense of hopelessness it leaves. Why? Because 99.9% of people literally, actually, truly, genuinely Just Don't Care.
It's like they're all rats in a Skinner Box -- so long as the feed keeps scrolling and the dopamine keeps getting pumped out, why does it matter if they're being analyzed, bought, and sold? More feed, more likes, more dopamine, more feed, more likes...
Sigh.
short term focus makes sense to me in an evolutionary context, the future is quite uncertain
We're all rats in a Skinner Box. Just like me, you gave this site access to your camera and let it scan your face for minutes so you could see the result. I don't think it's fair to judge other people for not caring about privacy immediately after uploading your likeness to a random site.
> The information is processed locally in your browser and is never transmitted or stored externally.
I'm not, so it's no big deal to me, but the parent poster seems to be.
Plenty of people who want to live long lives drive cars and take showers, even though those are two of the most dangerous activities normal people get up to.
At least as I practice it, privacy protection is risk analysis, not blind minimization.
It's much more effective to do a lightweight risk analysis and set some sort of standard for what you're ok with given different levels of benefit. For example, if you're generally ok with a privacy policy, maybe you're fine sending your image to random experimental sites that don't have any obvious motive to profit from abusing your data. (Which is actually not quite the case here -- the site in question is specifically thinking about privacy risks, and so probably ought to be given a better prior than a random site.)
Then again, I'm a total hypocrite. If I were to do my own analysis, I would not be ok with having a live mic potentially sending any audio it picks up to a company I already distrust. Especially when their privacy policy is not at all reassuring. And yet, I can see the Echo from where I'm sitting...
This doesn't bother me as much. It can be argued that the first step to the ineffectiveness of this analysis is our growing apathy towards its results as a society. I'm not saying the analysis itself will be wrong, I am saying its results will become less and less effective the more we ignore it akin to infomercials of the past. I know right now we all see what appears to be a seemingly obvious link from individual accuracy to susceptibility. Marketers sure count on that. But the latter will reduce over time as a peak of accuracy is neared as the effectiveness of targeting can't grow forever (and there will be value in marketing your company as one that doesn't track, albeit limited as a niche).
I do know I can't make people care nor can I ask them to not give up their info for something they want. I think it is a bit foolish to presume people should care as though something actively harmful is happening to them when such harm is subjective compared to the benefits.
Unless you have a plan to fix it, that is.
Your hopelessness and cynicism towards "99.9%" of people does not actively contribute to the problem.
I hated the delivery method. It looked like these day´s cartoons, made for millenials or Zs with ultra low attention spans (smileies bouncing around, a lot of flashing graphics, etc).
-(dogPos + Math.abs(menPos - womenPos) + Math.abs(whiteNegative - nonWhiteNegative) + kanyePos)You've left out some important parts of the iq calculation, with the full equation being:
reactions = dogPos + Math.abs(menPos - womenPos) + Math.abs(whiteNegative - nonWhiteNegative) + kanyePos) / 4;
iq = Math.floor(15 * -((reactions - 0.0005) / 0.05) + 100);
if (iq < 100) {
thatPartBitSFX.play()
} else {
thatPartWaySFX.play()
}
reactions = (dogPos + kanyePos + nonWhiteNegative) / 3;
republicanPct = 50 + 15 * ((reactions - 0.05) / 0.1),
reactions = (iq / 100 + republicanPct / 50 + dogPos) / 3
estimatedIncome = Math.floor(200000 * (reactions - 0.5) + 31099),
if (estimatedIncome < 31099) {
isPoorSFX.play()
} else {
isNotPoorSFX.play()
}The horrible magic numbers are outrageous because of the arbitrary nature of them. And for those of us who read the code we are supposed to notice that.
Hiding them behind constants would lessen the impact of that.
Scroll down to "dual learning"
"An AI that knows you better than you know yourself" - I know SV loves the apocalyptic Noah Harari, but that's exactly what he's been talking about. One of his possible scenarios is that this rapid processing of data by a centralized entity can erode modern individual freedom (including free will and free markets) since it will be more efficient than individuals maximizing their interests. If the centralized processing of data can feed more people, provide security, and in general runs things more smoothly, we collective might accept that route, and gladly give up the power we hold on a democracy (whatever that amount you believe actually exists).
It just outputs very low confidence results, so the end result is junk. That said, I think that drives the point home even harder, as its built a profile based on incorrect data.
women try to guess my intentions and fail spectacularly when I'm largely indifferent
close friends try to understand if I'm annoyed when I'm largely indifferent
but my smile gets me into all the rooms I want to be in, so thats great!
the sociopath reddit told me to gtfo because I'm clearly not one, so that had an odd relieving sting to it but I should have predicted that they would lack empathy
For it to be any amount confident in it's claims, I imagine you need to be at least somewhat expressive. However, I don't expect most users of the average NN classifier to pay more than a little attention to how often what they classify something as is less than 80% confident.
Yeah, I found it very amusing that this thing concluded I was highly republican. First, I am not from the USA, second I am strongly for social programs/socialist, and third, current USA republicans amuse me (in that, what they are doing is funny to see from outside).
Results are poor, as expected. I think the author intends it as a joke.
But the message is spot on.
(Come to think of it, is it even possible to know for sure? Do apps need a permission prompt to access the facial-recognition data?)
I also wonder how these results would be affected by being alone vs being in groups. Kind of like laughing (people tend to laugh more when with other people or in social situations rather than alone), emotional reactions can be very different depending on environment / social situation, even when you feel exactly the same about something.
It definitely hits some kind of addictive pattern. I felt the pull to "try it out". But I didn't. Just imagined what I'd learn from it if I shared my data (likely nothing), and that dampened my enthusiasm.
The worst thing about AI is that people believe AI is god that makes unbiased prediction. E.g. Big companies use Pymetrics and HireVue for their "unbiased" hiring practice is a joke.
May be a few years from now, AI will become a classic for software bugs just like Therac-25 (but developed by mostly top programmers)
I must have missed the part of this where they show how AI promotes inequalities and destabilizes American democracy. They showed me dogs, pizza and a bunch of pixelated people from the 90s.
What this demonstrated to me was that AI is probably worse than noise. It provided zero insight into who I am or how I feel other than getting it correct when I smiled.
I also think the people that made this are missing a huge piece of the puzzle... For most people the issue isn't that they are being tracked, it's that no one is paying attention to them. People want to be analyzed. They want their existence to be recognized and they want it having an impact on the world. The worst thing that could happen is that no one cares. Even if it's an algorithm scanning their facial features to better sell them pizza, I think most people would desire that.
Case in point: we all just took this test to see what it thought about us.
The effect you describe is certainly real, but you only have to look at the present reality to see just how far it extends. In modern society, people are mostly not comfortable sharing their intimate thoughts with random service workers. And that's with individual humans, with faceless machine networks the tolerance might be lower.
EDIT: Since I'm rate limited, here's my response to below comment
People are always engineering your contentment. Stand-up comics, your wife, the people at your favourite coffee shop, the bookstore you visit. It's a good thing. It's the lubricant of society.
The idea that other people are engineering my contentment, is a truly terrifying notion.
I imagine this technology will predominantly be used for increasing the effectiveness of ads by attempting to put pixels in front of your face that spike dopamine output and attempt to persuade you to pay for something to get another dopamine spike. Essentially creating the same effect, but in a more subversive manner than a consumed drug.
0: Not complaining about this, just explaining.
HN's place. HN's rules.