Slashdot Mirror
Researchers Fooled a Google AI Into Thinking a Rifle Was a Helicopter (wired.com)
An anonymous reader shares a Wired report: Algorithms, unlike humans, are susceptible to a specific type of problem called an "adversarial example." These are specially designed optical illusions that fool computers into doing things like mistake a picture of a panda for one of a gibbon. They can be images, sounds, or paragraphs of text. Think of them as hallucinations for algorithms. While a panda-gibbon mix-up may seem low stakes, an adversarial example could thwart the AI system that controls a self-driving car, for instance, causing it to mistake a stop sign for a speed limit one. They've already been used to beat other kinds of algorithms, like spam filters. Those adversarial examples are also much easier to create than was previously understood, according to research released Wednesday from MIT's Computer Science and Artificial Intelligence Laboratory. And not just under controlled conditions; the team reliably fooled Google's Cloud Vision API, a machine learning algorithm used in the real world today. For example, in November another team at MIT (with many of the same researchers) published a study demonstrating how Google's InceptionV3 image classifier could be duped into thinking that a 3-D-printed turtle was a rifle. In fact, researchers could manipulate the AI into thinking the turtle was any object they wanted.
Many humans are also easily fooled into thinking that this is just a plain brick wall:
http://cdn.playbuzz.com/cdn/d2...
Vandalizing street signs has been possible for a long time, just remove one street sign, and replace it with another. I don't think this has ever been a wide scale problem.
Myth. US prisons are not full up with people for marijuana convictions, especially not for simple possession.
Of the 750K annual US marijuana arrests:
About 40,000 inmates of state and federal prison have a current conviction involving marijuana, and about half of them are in for marijuana offenses alone; most of these were involved in distribution. Less than one percent are in for possession alone.
There are 2.2 million US prisoners at the state and federal level, so less than 2%. It's such a small % that the keepers of the keys (do they use keys anymore?) can keep their prisons full by delaying parole releases.
But yes, ethnicity still plays too large a role in sentencing, so you're not completely wrong.
Happiness in intelligent people is the rarest thing I know.
Ernest Hemingway
Here we see Trump Derangement Syndrome in full effect. A story completely unrelated to Trump, and yet the poster manages to shoehorn a spittle-flecked rant into the thread. AND it got modded up. That's collective derangement. Remember folks, these people losing their shit are the same ones who told us they were qualified to rule us because they were so educated and erudite. Would actual educated people be throwing temper tantrums and acting out in public like this?
OK, I resent that. I was comparing Trump's success at fooling 62,979,879 separate instances of the same natural intelligence to what these scientists achieved. If anything Trump's achievement is greater since each one of the 62,979,879 instances of the NI Trump fooled with his garbage data is a functionally quite distinct variation of the base NI whereas these scientists only managed to bamboozle a single variant of a much more primitive AI with much less variation from instance to instance. How is that unrelated? ... plus I have done nothing but heap praise on Mr. Trump for his skills at generating highly plausible garbage data. I find it most interesting from a purely scientific point of view that NIs are every bit as easy to fool by feeding them complete garbage data as artificial AIs are even though NIs are supposed to be more sophisticated. In fact Mr. Trump, with his political career, may have significantly advanced our understanding of how NIs deal with adversarial examples.
https://www.bleepingcomputer.c...
A stop sign will still look like a stop sign to you or me, but can be seen by the car's AI into seeing something totally different.
If you think I voted for Trump because of this post, you're wrong. I voted for Dr. Jill Stein of the Green Party. Again.
Yes but you have to keep in mind that the U.S. imprisons more of its population than Russia. And almost 8x the rate of most civilized countries.
Well, sure, if you're willing to go full on apples-to-oranges with that comparison to civilized countries.
Happiness in intelligent people is the rarest thing I know.
Ernest Hemingway
The key problem with AI, is its trust in in its sources. They havn't programmed in a silly algorithm yet. When kids are learning to process the world. Kids learn when things are in the wrong context then it is probably silly or just wrong. Even if it from a trusted source, a kid will laugh at their parent if they are saying something that is contradicting their view of the world. Such as when the parent is playing with the kid, they substitute a toy car for a doll, and play with the car like a doll. The child find this amusing because the context is all wrong. The AI algorithm seeing this, would just say this toy probably of usage has expanded to be used as a doll so it must be a doll. There is no questioning saying "no, that is not how you play with that toy". it will take the source as factual and just add it to its list.
If something is so important that you feel the need to post it on the internet... It probably isn't that important.
Do stupid crimes, pay stupid prices.
If anyone wants my sympathy because they got busted for drugs, tough luck. Drugs are illegal, and many state/city/federal prosecutors punish harshly for them. If drug users are too stupid to figure that out, that's their fault.
And, yes, I know several drug users, any one of which could end up in jail for years if busted. And I have said that to them,
If you think I voted for Trump because of this post, you're wrong. I voted for Dr. Jill Stein of the Green Party. Again.
All of these vision AIs are programmed backwards -- for convenience. This random object looks "more like" a speed limit sign and "less like" a stop sign. Great. No body cares about how much "like" something another something appears.
You can ask any 10-year old. A stop sign is a red octagon. Any 16-year old will say it also has a white border, and white lettering in the middle. Any experienced driver will add that it appears at some sort of intersection, obstruction, or event, alongside a narrow road.
Now, if you see a red octagon, and you stop, and it turns out to be a giant lollipop, then that's good. Because a giant lollipop on the road is absolutely acting as a stop sign.
If something isn't a red octagon, then it's definitely not a stop sign.
The problem here is that google's vision AI doesn't identify an sight according to what defines a stop sign -- a red octagon on the side of the road. That's because it's highly stupid.
And the question really comes down to something much simpler. If I put a big square sign on the side of the road, blue, with yellow lettering, that says "please pause, thank you", will google treat it as a stop sign? Good bet that any driver who sees it (and can certainly be forgiven for not noticing it) will stop.
Conversely, on a highway, at 120kph, if I put a real stop sign on the side of the road, will google treat it as a stop sign? No human driver is going to slam on the brakes.
Google's not thinking. Therefore, it ain't an AI. It surely "looks like" an AI, but it's not an AI. It uses collected intelligence to determine what the object is, but it doesn't use its own intelligence to make decisions. It doesn't make decisions at all.
Show me a vision system that can take any photograph of any road, and decide whether or not it should stop the car. Doesn't need to be right or wrong, correct or incorrect, it just needs to make a decision, reliably, that makes sense. See, if it can do that, "reliably", then we can change the signs for them. We chose the signs for us for a reason. Humans see red first, so stop signs are red. If machines have trouble with octagons, and love purple, then we can give them that instead. Dual signage is common in multi-lingual communities.
But these shitty AI systems are much worse. They don't even make their classifications reliably -- because the more data they collect, the most they distract themselves. So a guaranteed "this is a stop sign, 100%" can change a year later, as it "learns", such that the very same stop sign is now only 80%. There's no fortification. There's no stubbornness. That's a problem.
I do not think the system was actually "fooled" It was taught the wrong thing.
Well on principle the solution is simple then: only teach the system the right thing. Just as in programming you can avoid bugs by not making mistakes.
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
We have had wars over less than what is "the right thing". Please don't.
Don't fight for your country, if your country does not fight for you.
You make several valid points. The biggest rebuttal about them is that the countries you mention are mostly homogenous populations, without the racial history that defined the US. The biggest problem (imho) today is that the US has a large population that is being told every day that they have to "stick it to the man". Modern black culture in America is its own worst problem, and many who try to escape it are punished by that same culture for "being white".
Poor immigrants come in from Asia every year. We don't see the problems in the poor Asian communities that we see in the poor black communities. So it isn't simply a "non-white" problem. It is a cultural problem withing the modern day black community. Part of the blame can certainly be placed on racism, but that once "whites only" was outlawed, racism is more of a crutch than actual reason for someone being held back.
For many of the programs that worked in other areas, we have seen them attempted here in the US. They fail within a few years because
1. It was a political move, and ignored after the election
2. Only a dozen people were successfully helped, out of thousands/millions
3. Blacks actively attacked the blacks that participated
4. Clinics/offices in the neighborhoods served were vandalized or burglarized
5. Corruption, bribes, slush funds, rather than real aid
6. Too many people allow all of the above to happen
If you think I voted for Trump because of this post, you're wrong. I voted for Dr. Jill Stein of the Green Party. Again.
If and when self-driving cars really become a thing, vandalism of street signs will probably have to be elevated to a felony with a mandatory minimums, even if no one gets hurt. It'll also have to be something where minors can be charged as adults because they're the ones who probably do the majority of it, and you know there will be teens who'll think it's funny to cause a 10 car pile up.
As others here note, traffic disruption pranks aren't a big problem now even though stealing signs, or introducing obstacles is already possible.
No defense against attempts to disrupt traffic is going to cover all cases, but an excellent one already implementable for all self-driving systems should be obvious.
Self-driving cars aren't navigating a blind road grid, they already have virtually complete maps of the entire road system. Give each car a database of the location of all signs in existence. Humans need signs to inform them of the traffic law in effect at an intersection (for example) computerized systems do not.
Starships were meant to fly, Hands up and touch the sky - Nicky Minaj
Optical illusions work because our visual system takes certain shortcuts to reduce the amount of processing needed to identify what it is we're looking at. e.g. We assume diagonal lines are 3-dimensional, leading to errors when we view a 2D object with diagonal lines. The only information actually provided was the horizontal line + 2 diagonal lines. Our brains extrapolated the nonexistent 3D nature of the object to create the error. (The top line looks like the the edge of a box viewed from the inside, so our brain concludes the line is further away and thus bigger than it appears; the bottom line looks like the edge of a box viewed from the outside, so our brain concludes the line is closer and thus smaller than it appears.) Likewise, the computer vision AI makes the turtle/rifle error because it's extrapolating from its very limited information subset to determine if the object is a turtle or a rifle.
These sorts of errors disappear as you add more information, thus reducing the amount of extrapolation needed. I was driving on a rural highway at night when suddenly it seemed like the road was twisting and warping. This went on for about 10 seconds until I moved into an area with fewer trees, and I realized what I thought were billboards in the distance were actually boxcars on a moving train. My brain had been assuming they were fixed points in space, when in fact they were moving. So initially it erroneously concluded the billboards were static and the road was warping, but the moment I recognized them as boxcars my brain correctly realized the "billboards" were moving and the road was static.
So in these early stages of visual AI, we're going to encounter a lot of these errors. But as the AI becomes more sophisticated and able to take into account more contextual information, these errors will begin to disappear. They probably won't disappear entirely, because you can only glean so much information from a static photo. But for real-life applications like security, the turtle/rifle error is highly unlikely to happen once the AI starts comparing the questionable object in multiple frames in a video instead of a single frame, or starts comparing it from multiple viewpoints provided by multiple cameras.
It might make more sense, but that is not the way it is. I think we have found your problem. You think the world will make sense, with a thought process like that the next thing you will do is conclude that someone must have designed it that way.
The truth is that all men having power ought to be mistrusted. James Madison
Actually, if you adjust for being raised in a single parent household vs being raised by both parents, the difference in conviction rates between whites and blacks disappears (or, at least drops to the margin of error of studies on the subject). The same thing happens to poverty rates.
The truth is that all men having power ought to be mistrusted. James Madison