Slashdot Mirror
Microsoft's New AI Mistakenly Identifies Photos, Ignores Hitler (mashable.com)
An anonymous reader writes: Microsoft's newest online AI, CaptionBot, tries to identify what's in an uploaded photo, using two recognition APIs recently released by Microsoft Cognitive Services for app developers-- "Computer Vision" and "Emotion". But while Microsoft brags that their AI "can understand thousands of objects, as well as the relationships between them," bloggers are also sharing funny examples of CaptionBot's many mistakes. While it correctly identified Bea Arthur, Ozzy Osbourne and Joan Jett, and a movie poster with Arnold Schwarzenegger, it mistakenly identified Gene Simmons of KISS as "a woman in a red jacket...sitting on a motorcycle," described a wedding dress as "a cat wearing a tie," mistook Michelle Obama for a cellphone, and described one man's Twitter avatar as "a close up of two giraffes near a tree."
But CNNMoney reports that the AI is apparently programmed to ignore all images of Hitler and other Nazi symbolism (as well as Osama bin Laden), reporting that Microsoft's AI "often came back with 'I really can't describe the picture' and a confused emoji. It did, however, identify other Nazi leaders like Joseph Mengele and Joseph Goebbels."
But CNNMoney reports that the AI is apparently programmed to ignore all images of Hitler and other Nazi symbolism (as well as Osama bin Laden), reporting that Microsoft's AI "often came back with 'I really can't describe the picture' and a confused emoji. It did, however, identify other Nazi leaders like Joseph Mengele and Joseph Goebbels."
They don't want another nazi-bot
But then the FBI came along, and then all the smart phones were like, "we were totally wrong. everything is fine. we're sorry. government is good."
rewriting history since 2109
They should call it artificial dogma rather than intelligence.
Microsoft's AI keeps embarrassing them. It's like they thought their corporate image problem from being a ham-handed OS monopoly wasn't big enough: they needed to automate gaffes.
Table-ized A.I.
As an online discussion grows longer, the probability of a comparison involving Nazis or Hitler approaches 1.
Interestingly, the "wedding dress" from the summary is not a wedding dress at all. In fact it's the famous black/blue or white/gold dress. The reason that's interesting is because the 'human' in charge of complaining that the computer can't recognize a simple image....can't recognize a simple image.
Additionally, that whole 'dress color' kerfuffle shows that image recognition can be a difficult task even for the human brain...which has been specifically designed and built over thousands of years to do that very thing.
I think the most important piece of information here is that AI just isn't ready for the big time yet. People are going to do and say all kinds of fucked up and bizarre things. People will try to have sex with anything. They'll try to convince their AI assistant to support genocide. They'll demand that it pretend to agree with them about things like that. They'll ask for information that's not available, they'll cuss and scream, they'll talk about things that seem completely off-topic. People will use puns, innuendo, and vague references. They'll yell insults, start stupid arguments, and lie through their teeth even when it only hurts themselves. And yes they will talk about Hitler. An AI that can properly handle all of this and not go off the deep end is a full-fledged strong AI.
We're not there yet but this effort by Microsoft is, IMHO, as smart as a mouse. And with geometric progress that means the real thing won't be long now.
If video games influenced behavior the Pac Man generation would be eating pills and running away from their problems.
The most ridiculous might be what Microsoft's AI describes an "an operating system suitable for mission critical servers". Or maybe that was Microsoft Marketing, not Microsoft AI. Either way.
This is not Missed Connections?
I'm reminded that about half of Slashdotters are afraid that AI like this will put them out of a job soon. The other half of Slashdotters can tell the difference between a cell phone and the first lady, so they won't be replaced by Microsoft software.
On the other hand, 15% of Slashdot readers can't tell the difference between Obama and Hitler, with this AI can do so.
http://ignorehitler.tumblr.com...
Inheritance is the sincerest form of nepotism.
Does it identify the destruction of the Twin Towers and Building 7 as a controlled demolition ?
Microsoft's AI keeps embarrassing them. It's like they thought their corporate image problem from being a ham-handed OS monopoly wasn't big enough: they needed to automate gaffes.
It is trivially easy to get a instant mod-up on Slashdot by pointing to the Microsoft's AI's occasional mistakes and not its successes. But most of the time Microsoft's AI seems to be getting it right. If you have something better, put it up where we can see it.
They don't have any problem identifying photos of Hitler as Hitler. The problem is false positives: If the software mistook the photo of some living person as Hitler, and that was somehow published, that person would not be happy, and might start a lawsuit.
Problem is easily solved by telling the software "if you think it is Hitler, you say you don't recognise it". There was a case a while ago where some photo analysis software mistook a woman for a gorilla. Highly embarrassing for everyone involved.
I would think that software makers would nowadays add precautions to make particularly embarrassing mistakes less likely. (Mistaking a gorilla for a woman is no big deal, the other way round it's very bad).
Did we learn nothing from the time we made HAL lie?
--
I think so Brain. But why do I have to wear this itchy & scratchy toothbrush on my upper lip?
One wonders which caption it would put on goatse?
Well - it recognised Joseph Mengele, although describing him as a "Nazi leader" is ridiculous.
Mielipiteet omiani - Opinions personal, facts suspect.
it was just perfect on MS provided pictures and was zero on pictures i provided.
Upload picture of floppy it can t describe it.
Upload identity picture of me, it s a man holding a remote control
Upload a photo of the hearth , it s a close ip of a wave
https://www.captionbot.ai/
I am not really sure, but I think it's a small off-duty Czechoslovakian traffic warden.
I can't really describe this picture- but i do know it did nothing wrong!
If it was a pic of you and the AI would say "Oh, that's Adolf", how'd you feel?
We used to have a Bill of Rights. Now, with the rights gone, all we have left is the bill.
As soon as this bot goes live, will we only get to see Hitler pics to solve CAPTChAs so the botters don't get in?
We used to have a Bill of Rights. Now, with the rights gone, all we have left is the bill.
All kinds of AI have teething pains, during which the problems are obvious and comical (the Apple Newton's handwriting recognition being a case in point). At the same time, the achievements of modern AI are amazing--but also troubling.
When I compare AI as envisioned in the 1950s--Isaac Asimov's Multivac, or his robots, perhaps--the assumption was that AI would be closely similar to human intelligence. For example, it was implicit that robots would answer questions by actually understanding them. What we are seeing today evokes an analogy with technologies like the sewing machine. Early efforts attempted to sew the same way humans did, and failed. Singer's brilliant idea was a method of using thread to fasten two pieces of cloth that did not resemble human sewing or even use the same stitch.
A Google search is within shooting distance of Multivac. You type in a question and you get a useful answer. The interesting thing is that most modern AI is shoddy. It goes halfway. It gives you something that's inaccurate, yet useful. But the key thing is that you are expected to use your human intelligence to get the rest of the way and correct mistakes. In the case of Google, you do this by looking at a ten or a hundred search results, for example--and reformulating the question if you don't get the right answer.
Perhaps one of the things that early AI pioneers missed is that modern AI relies more on having huge databases of information than would have even been imaginable in the 1950s and 1960s, and less on AI actually mimicking human intelligence.
This is not a problem when it is all open, the AI is offering you something to look at and not making decisions for you, and it is all in the nature of help or suggestions rather than direct action.
It becomes far more serious when it is happening behind the scenes--when AI is deciding whether you get a loan, or pass an essay test on an exam, or get onto a terrorist watchlist.
"How to Do Nothing," kids activities, back in print!
Ignoring is not actually a term that I'd use when describing Nazis. In fact, if they'd just ignored those they hated then there probably wouldn't have needed to be a giant FUCKING WAR because of their behavior. Slaughtering millions and bombing other people's property into rubble is not exactly "ignoring." Unless, of course, you've got a very different definition of ignoring than I do.
"So long and thanks for all the fish."
Better than being identified as a phone or giraffe.
Cut him some slack. He's upset the none of the preachers wanted to molest him. It has jaded him for life.
You don't suspect that has already happened? I mean look at some of these posts explaining the flaws or any MS blunder.
Even if we don't like what he did, Hitler did actually exist and is a significant character in world history,
If we choose to ignore history we're doomed to make the same mistakes again.
Nah, twice in a row is bad style.
We used to have a Bill of Rights. Now, with the rights gone, all we have left is the bill.
As soon as I heard that someone's avatar was described as being two giraffes, I knew it was going to be in black and white. As far as I can tell, their algorithm thinks that any greyscale image includes two giraffes. A rorschach test image, an art piece with a stylised tree, a black and white MS Paint picture of a stick-man Dumbledore, everything I could find got described as two giraffes (often in a "fenced-off area").
IBM - Watson, a computer that can win at Jeopardy
Google - AlphaGo, a computer that wins at Go
Microsoft - Tay, a racist chatbot
Jeopardy is a trivia game.
Key words and phrases to which you respond with a factoid. To be fun and playable for the audience the boundaries of this "universe" have to be quite small.
Go is a game which is played with perfect information and clearly defined rules. It is a fascinating problem in its own right but it is not the same problem as recognizing a face or an object in a purely arbitrary setting.
Joseph Mengele was not a Nazi leader. He was a member of the Schutzstaffel, and a registered physician at the Auschwitz death camp. He was infamous and notorious for his sadistic behaviour, which included sickening human experiments involving sewing live people together (in an attempt to recreate conjoined twins), injecting chemicals into victims' eyeballs (in an attempt to learn about eye colouration), murdering people with chloroform for the purely sake of dissecting them, and other brutal pseudoscientific activities. To the best of my knowledge, he was never considered among the political elite (i.e. the 'leaders') of the Nazi party.
He was just a low-level "researcher" in a KZ that got notorious because he killed so many people in so many different gruesome ways (and enough of his subjects still survived the ordeal to tell the story).
His superior back in Berlin more or less continued his career after the war - mostly because he systematically destroyed most documents that could proof a connection with the notorious experiments (once it was obvious that the war wasn't going to end well for Germany) and because Mengele himself had fled to South America,drawing all the attention onto himself.
Windows 2000 - from the guys who brought us edlin