Slashdot Mirror
Researchers Create 'Psychedelic' Stickers That Confuse AI Image Recognition (techcrunch.com)
"Researchers at Google were able to create little stickers with 'psychedelic'-looking patterns on them that could trick computer AI image-classifying algorithms into mis-classifying images of objects that it would normally be able to recognize," writes amxcoder:
The patterned stickers work by tricking the image recognition algorithm into focusing on, and studying, the little pattern on the small sticker -- and ignoring the rest of the image, including the actual object in the picture... The images on the stickers were created by the researchers using knowledge of features and shapes, patterns, and colors that the image recognition algorithms look for and focus on.
These stickers were created so that the algorithm finds them 'more interesting' than the rest of the image and will focus most of it's attention on analyzing the pattern, while giving the rest of the image content a lower importance, thus ignoring it or confusing it.
The technique "works in the real world, and can be disguised as an innocuous sticker," note the researchers -- describing them as "targeted adversarial image patches."
These stickers were created so that the algorithm finds them 'more interesting' than the rest of the image and will focus most of it's attention on analyzing the pattern, while giving the rest of the image content a lower importance, thus ignoring it or confusing it.
The technique "works in the real world, and can be disguised as an innocuous sticker," note the researchers -- describing them as "targeted adversarial image patches."
It looks as if the AI is concentrating on the area with the most detail, even though it is not really relevant. I've seen similar, ummmm, distractions confuse AI. For example, disguising a stop sign so that a self-driving car is confused.
Oh no! Our spying may be tampered with!
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
1. Add stickers to images.
2. Retrain network
3. Stickers useless.
Amazon will be selling hats and scarves with psychedelic looking patterns on them.
You live and learn, or you don't learn much.
Perhaps he left to attend his brother's funeral?
Remember the "worlds ugliest t-shirt" in one of William Gibson's novels? All cameras in that book's world were compelled by their firmware to fill image of the wearer of that suit with background. One could laugh at such a notion except ....scanners won't do banknotes
"I thought what I'd do was I'd pretend I was one of those deaf-mutes"
Reminds me of Ghost in the Shell's Laughing Man calling card... His sticker would appear over people's faces in VR if they were infected.
Would one of these stickers on the bumper of my car defeat the automated license plate readers?
Just waiting for manufacturers to start selling $10 stickers, shirts, hats, backpacks, luggage tags etc.
When's the IPO?
When the robots take over our jobs and then decide we aren't needed, we'll just get them addicted to these stickers. They'll soon get bored with theirs and go looking to trade each other for new ones. Then they'll begin their own industry of trippy stickers so they can get a better high. All day they'll sit and run their batteries dry. RIP to the bots that get stuck in a while loop.
With a similar enough network or access to the targeted network, simply create a network that learns to fool the other one. Loosely like two computers playing chess but more like a spam generator to defeat filters.
Adversarial network learning... just not an official use of it... The solution is to add this kind of learning to the network... except it won't be fool proof until the network is quite good; since the adversary could have as many variations of attack as the classifier has in recognition.
If you created the adversarial network used to train it, you could leave INTENTIONAL holes for future exploitation. Even going so far as to purposely train in holes if you had that kind of access. It's not like anybody is going to spot your code in the AI -- only the training setup... which could be long gone after years of training... In the future, I would expect to have VALUE in AI training whereby the cost of "reboot" would be quite significant... finding bad training data over millions of samples and years of experience could be difficult and who's to say all that would be retained? You take the resulting network from last week and retrain from that point-- you'd not go back years ago and restart. I'm talking way out... because AI is so simple now you can just archive all input data... maybe by that point we can still archive it all and learning hardware will be faster... anyhow, it makes for interesting Sci-Fi possibilities even if it may never become an issue (even if it doesn't, there would still be a cost involved in retraining from scratch.)
Democracy Now! - uncensored, anti-establishment news
Correct. Now all we need to do is gather every possible psychedelic sticker possible to begin our training.
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
If the image was a true unknown, then who would know if the sticker or the banana is more "key?". Just need a time constraint for AI to write off a weird thing as "weird thing" and move on to the next pattern in the image.
Now that I think about it, I would be curious how the AI would handle a jumble photo, and be able to identify all the stuff in the picture?
sigs are for losers (except to point out that sigs are for losers)
Is not as easily fooled as this pattern matching NN grossly incorrectly hyped as Artificial intelligence. Just saying - hype is hype no matter how much you want to believe you've got the next big thing and innovation (and in this case, NN research and pattern matching work go WAY back).
Why guess when you can know? Measure!
Put up a topologist's-sine-curve-weighted gradient. If there is an AI which can discern it, it's either not refined enough or it's the next step. I guarantee that no neural net will ever handle it.
Any guest worker system is indistinguishable from indentured servitude.
They seem to mess Bender up a bit.
It must have been something you assimilated. . . .
Our "real" human visual algorithms are distracted by bright, shiny objects in a similar way. It's not just AI that can be fooled.
1. The AI assumes that it always sees only one object.
2. How can it classify this sticker as a toaster? It should be classified as unknown. I think they cheat by assuming that every image can be classified
A human looks at that picture, sees the banana and "thing" are sitting on a flat surface, and decides they must be about the same distance so their size in their picture is their actual scale. The banana is a lot bigger, so the human decides it is more important than the "thing".
An AI looks at that picture, sees the banana and "thing", but crucially doesn't estimate distance. Since the "thing" has a lot more detail the AI decides it's must be further away, and its greater detail means its the more important part of the picture, and the banana is just fluff in the background. And it gets lost trying to analyze the "thing".
It will be useful when we're trying to fight SkyNet during the inevitable upcoming robot apocalypse.
Hire a Linux system administrator, systems engineer,
While not exactly the same thing, in one of William Gibson's recent trilogies the characters wore clothing with specific patterns that were designed to render them invisible to surveillance cameras. The basic premise was that the even though the cameras recorded them, the computers monitoring the cameras did not realize that there were people in the images.
Can we expect to see this appearing as part of Captchas, then?
Meanwhile, Lisa Frank sticker sets see a huge sales growth!
So, they figured out how stoners' brains work.
Have gnu, will travel.
I was hoping for something we could put on a cheek that would thwart facial recognition software. Or at least make me look like somebody better looking. (No, I wasn't looking for a full-face mask.)
There's no time like the present. Well, the past used to be.