Slashdot Mirror
A Neural Network Can Learn To Recognize the World It Sees Into Concepts (technologyreview.com)
An anonymous reader quotes a report from MIT Technology Review: As good as they are at causing mischief, researchers from the MIT-IBM Watson AI Lab realized GANs, or generative adversarial networks, are also a powerful tool: because they paint what they're "thinking," they could give humans insight into how neural networks learn and reason. [T]he researchers began probing a GAN's learning mechanics by feeding it various photos of scenery -- trees, grass, buildings, and sky. They wanted to see whether it would learn to organize the pixels into sensible groups without being explicitly told how. Stunningly, over time, it did. By turning "on" and "off" various "neurons" and asking the GAN to paint what it thought, the researchers found distinct neuron clusters that had learned to represent a tree, for example. Other clusters represented grass, while still others represented walls or doors. In other words, it had managed to group tree pixels with tree pixels and door pixels with door pixels regardless of how these objects changed color from photo to photo in the training set.
Not only that, but the GAN seemed to know what kind of door to paint depending on the type of wall pictured in an image. It would paint a Georgian-style door on a brick building with Georgian architecture, or a stone door on a Gothic building. It also refused to paint any doors on a piece of sky. Without being told, the GAN had somehow grasped certain unspoken truths about the world. Being able to identify which clusters correspond to which concepts makes it possible to control the neural network's output. The team has now released an app called GANpaint that turns this newfound ability into an artistic tool. It allows you to turn on specific neuron clusters to paint scenes of buildings in grassy fields with lots of doors. Beyond its silliness as a playful outlet, it also speaks to the greater potential of this research.
Not only that, but the GAN seemed to know what kind of door to paint depending on the type of wall pictured in an image. It would paint a Georgian-style door on a brick building with Georgian architecture, or a stone door on a Gothic building. It also refused to paint any doors on a piece of sky. Without being told, the GAN had somehow grasped certain unspoken truths about the world. Being able to identify which clusters correspond to which concepts makes it possible to control the neural network's output. The team has now released an app called GANpaint that turns this newfound ability into an artistic tool. It allows you to turn on specific neuron clusters to paint scenes of buildings in grassy fields with lots of doors. Beyond its silliness as a playful outlet, it also speaks to the greater potential of this research.
Does it go berserk and kill 5 billion people?
I think these guys are building a matrix...
I remember seeing this and playing with GANPaint last November. It was on HackerNews, Twitter,... The article reads, "The team has now released an app called GANPaint." Now? The tweet right above this text, announcing GANPaint, is from November 27, 2018. So...is there something new, or is MIT just now getting the memo?
"Alexa, draw my hands larger on the news."
Table-ized A.I.
Another decade of the AI winter avoided.
The AI can now "learn", do "thinking" and "see".
Funding secured.
Domestic spying is now "Benign Information Gathering"
Don't get me started. Computer mice don't even look real. Computer scientists are so dumb they think trees grow with their branches and leaves pointed downward. I would like to see someone smoke a hash function, you can't!
Everyone in the field knows this. They distinguish neural networks from neuronal networks, which are intentionally more faithful to biology.
Nevertheless there is some commonality between neural and neuronal networks in overall aspects of computation. The first big book already had the right name: Parallel Distributed Processing.
So it statistically correlated randomly-grouped information over millions of trials?
Still not AI. Still just statistics. Bad statistics. With complete lack of inference. With shady, if not downright dishonest, assertions made about its capabilities.
You did what "AI" has had done to it for decades now... flip a bit to indicate success in some fashion, and throw millions of trials at it until it trains itself to activate "bit" more than a handful of times.
It could be flipping because it's majority green. Because the top-left pixel is green. Because there's watermark on the image. Because some frequency curve (if you have those, it's highly unlikely to form those itself in even a billion attempts / generations / evolutions / trainings) hits on a certain colour.
The fact is: You have NO idea what it's correlating on. It's almost superstition on the part of the AI (if it wasn't completely lacking in any intelligence).
Did you know that if you feed a pigeon in a box at random times it starts to associate feeding time with whatever it happened to be doing, and so repeat that? Whether that's bobbing its head, pecking at the floor, or looking a certain direction.
It then spends most of its time trying to replicate that convinced that it's "just not doing it quite right", like someone with a superstition about their team winning because they were wearing their lucky underpants - no amount of negative correlation will convince them they are wrong and get them to change their ways.
And that's exactly the problem with "AI" / neural networks. Of course you can train them to a statistical correlation - you know why? Because you're eliminating / training out / not breeding from those that don't correlate somewhat. It doesn't mean that what they are working from has anything to do with what you were after. And, most importantly, it does not mean you can trust them further on new data, nor that you can "untrain" them when they get it wrong, nor that you can perpetually improve them by more and more training.
All that happens is that it plateaus before it ever really gets useful (usually within the range of a PhD study - write your thesis up quick!), people release rubbishy apps "to show what it can do" and then it's never touched again because it can't be used for anything else and isn't particularly good or useful at what it does do.
We don't have AI, stop trying to pretend that we do. When you get a machine that can infer, that can actually reason its answer (not just "well it matches shape 22%, colour 17% and overall pattern roughness 19.4%", but "I can see branches here, here and here. They are connected. The connection grow and increase in width. The thickness part, which looks trunk-like, ends in a solid base which resembles soil", etc.)
Until then, this is all just a waste of time, and heuristics (YOU told it when it got the tree right).
MIT PR department overhypes absolutely everything to the point of making it next to impossible to understand what is actually new and valuable research. Well, I guess that keeps the money flowing in, but sometimes it would be nice if they wouldn't appear to mostly act as an obfuscation layer between their researchers and the reader.
The timing is interesting as even journalists are starting to turn against what passes for 'AI' these days...I forsee an AI winter next year..
"So you're saying strong AI has potential? Sounds good. When though."
None dare call it strong AI, that's all. Pitch it as an approach to extended versions of the same sort of problems that narrow AI is solving, and you will partake of the same rich funding as narrow AI.
I would like to see someone smoke a hash function, you can't!
No, but you can let the magic smoke out of something that runs it. With enough acrid solder smoke floating around you won't know the difference.
CPU instruction HCF: Halt and Catch Fire. Link.
This was a TV Show?!? I didn't know.
If the universe is someone's simulation -- does that mean the stars are just stuck pixels?
I think our 'intelligence' is just more layers of pattern recognition. And, by patterns, I don't mean just what our 5 senses bring in. The brain monitors additional inputs. We construe those as emotions, feelings, guesses, etc.
Brain or not, it works in the real world. It's been a standard technique in particle physics since the early 1990's, and has been used to make real discoveries. You believe it's not thinking? Fine. But that just means it's increasingly possible to do these very sophisticated tasks without thought, which could hitherto only be done with thought. That's big news, not snake oil.
It can learn a style like georgian, that is still impressive.
People keep saying this but I've written assembly for a few different machines and I've never actually seen a HCF opcode in any of them. Is this some sort of nerd urban legend?
You can even sometimes tell what is frying by the smell and color of the magic smoke. Not much smoke but a highly acrid smell? Likely your typical electrolytic capacitors venting - often smelled from some cheap dimmable LED bulbs that fail after only a few hours. It's not common knowledge but they hide all the colors in any LED, you simply need to run massive current through it and enjoy the short lived show. My coworker could tell what basic type of resistor was over heating by the smell, almost as awesome as my old shop teacher who could taste oil weight. My personal favorite is the bright pink smoke you get from frying a particular brand of power mosfets. I still find it funny one of the best ways to spot problems on boards is to forgo any knowledge of current, voltage or electricity in general and simply use an IR camera to spot what's hot that shouldn't be.
"Recognize the World It Sees Into Concepts" ?
Either the editor is illiterate and made a gross grammar error, or the headline should read:
"Reorganize the World It Sees Into Concepts"
In defense of the editor, they usually just copy stuff from a source without reading it. Perhaps the source was wrong this time.
...omphaloskepsis often...
You dummy, they were Australian scientists!