Skynet Surfs For Cat Pics

Skynet Surfs For Cat Pics

image

Google's massive, artificial neural network has learned to love the Lolcats.

The big brains at Google recently moved humanity significantly closer to extinction by creating a 16,000 CPU neural network with a billion connections and then turning it loose on YouTube. But as it turns out, the AI is less interested in heralding doom for all organic life than it is in the endlessly amusing antics of kittycats.

Three days, ten million random YouTube thumbnails and one "deep learning" algorithm later, the G-brain was able to pick out pictures of cats from a list of 20,000 different items with nearly 75 percent accuracy. Less amusingly, it achieved 81.7 percent accuracy in detecting human faces and 76.7 percent accuracy in identifying human body parts. What makes this achievement remarkable is that the machine wasn't given any information to help it figure out what was what; instead, it figured things out entirely on its own.

"We never told it during the training, 'This is a cat'," said Dr. Jeff Dean, the Google fellow who led the study. "It basically invented the concept of a cat."

"Contrary to what appears to be a widely-held intuition, our experimental results reveal that it is possible to train a face detector without having to label images as containing a face or not. Control experiments show that this feature detector is robust not only to translation but also to scaling and out-of-plane rotation," the more scientifically-precise "Building High-Level Features Using Large Scale Unsupervised Learning" paper explains. "We also find that the same network is sensitive to other high-level concepts such as cat faces and human bodies."

And while even the massive G-brain is "dwarfed" by real human brains, David Bader, the executive director of high-performance computing at the Georgia Tech College of Computing, said that more accurate modeling is coming quickly. "The Stanford/Google paper pushes the envelope on the size and scale of neural networks by an order of magnitude over previous efforts," he said. "The scale of modeling the full human visual cortex may be within reach before the end of the decade."

The success of the project has resulted in its relocation from Google's "X lab" to its primary facility, a move that has almost certainly brought the end of days several steps closer. Even so, Dr. Andrew Ng, Dean's co-lead on the project, thinks that truly self-teaching machines are still a long way off. "It'd be fantastic if it turns out that all we need to do is take current algorithms and run them bigger, but my gut feeling is that we still don't quite have the right algorithm yet," he said.

Then again, if it digs cat videos on the internet, how far off can it be?

Source: New York Times

(photo)

Permalink

Next up: G-brain finds the doritos. Productivity drops markedly.

... They wanted to make something intelligent
... They set it loose on youtube?

I don't know what to say, I can only imagine that within just a few days it will qualify as a high functioning autistic atheist, that thinks that Ron Paul is the answer to al 'Murica's problems, loves misbehaving cats dogs and babies and is convinced that XxXI1337Sk1775IXxX is the best MW3 player of all time!!!111one

Oh and the whole "we just created a fucking learning robotic brain" thing that Google just did? Real smart, like intelligent smart in that it must have required a lot of brainpower to do, sarcastic smart in that you created a learning robotic brain and released it on youtube

If you are going to set a learning artificial mind loose, maybe start with news blogs with comments disabled and see if it gains a capacity for understanding bias and the need for factual evaluation of the information that is presented to you, rather than, you know... a site with millions of end of the world plans, combat instructional videos, language training techniques, Americans and epic meal time... there is no possible way that this could end well.

Also... cat faces and human bodies? it found that part of youtube didn't it?

It found cats? Are you sure that's all it did? What if it identified it's new master and assumes that cats want to enslave humanity?
Be afraid, be very afraid!

It scares me that first the computer learned people and then it learned cats. I don't think this was just happenstance. It first identified humans, the enemy, it next identified our weakness, adorable cats... Cute bundles of kitty will be the herald of mankind's destruction.

When we can read about advances in computer technology without belaboring the end of days, THEN we'll be doomed. Keep up the good work.

It's also fair to say that bombarding someonething with ten million things on Youtube over a mere three days is probably an excellent way to convince it that we should be wiped out.

Good going, Science. Hopefully my kitty will speak up for me when the robot revolution comes knocking at my door.

So is it sentient or did it just do as it was programmed to do?
No jokes please this is a serious question.

Coulda been worse, they could have let it loose on 4chan or 9gag. I mean Youtube ain't much better, but imagine if it learnt humanity from those assholes.

Captcha: We go forwards. Yes, captcha, into the darkness.....

What a coincidence, I just received a study book on machine learning. (I am studying AI.)

rolfwesselius:
So is it sentient or did it just do as it was programmed to do?
No jokes please this is a serious question.

It just did as it was programmed to do. It's an advanced datamining algorithm.

Funny thing is, they checked it's performance using images off imagenet and had an accuracy of 15%. But it's accuracy for human faces and kitties was much higher because they trained it using video from youtube. (Basically it's a roundabout way of saying youtube has lots of cats.)

LGC Pominator:
... They wanted to make something intelligent
... They set it loose on youtube?

I don't know what to say, I can only imagine that within just a few days it will qualify as a high functioning autistic atheist, that thinks that Ron Paul is the answer to al 'Murica's problems, loves misbehaving cats dogs and babies and is convinced that XxXI1337Sk1775IXxX is the best MW3 player of all time!!!111one

Oh and the whole "we just created a fucking learning robotic brain" thing that Google just did? Real smart, like intelligent smart in that it must have required a lot of brainpower to do, sarcastic smart in that you created a learning robotic brain and released it on youtube

If you are going to set a learning artificial mind loose, maybe start with news blogs with comments disabled and see if it gains a capacity for understanding bias and the need for factual evaluation of the information that is presented to you, rather than, you know... a site with millions of end of the world plans, combat instructional videos, language training techniques, Americans and epic meal time... there is no possible way that this could end well.

Also... cat faces and human bodies? it found that part of youtube didn't it?

There's no need to get get your panties in a twist. It wasn't an everything-learning robot that was "set loose" on YouTube. It was a learning algorithm that was fed a couple hundred thousand random thumbnails of YouTube videos. The fact that it learned about cats and humans means just two things:
1) There are relatively a lot of videos which feature humans and cats on YouTube
2) It is an impressive learning algorithm

This is really cool technology though. Machine learning algorithms can already learn to do most classification tasks from raw data (e.g. pixels) in theory, but in practice it was almost always necessary for people to manually preprocess that data to extract more meaningful features (e.g. edges or corners). Coming up with new, better features is quite the art. And that is basically what these deep learning technologies do: find good features in the data (so that scientists don't need to). One really high level feature is "cat face". It is still very computationally expensive, but obviously it is moving into the realm of possibilities.

By the way, it is also nothing like the brain. Researchers like to look at the brain for inspiration, and when they come up with something new, they will say that their technology is now more brain-like than previously. And that is correct, but brains are still very different. Judging from the Google paper, they are just using a small handful of relatively simple[1] ideas/algorithms on a huge scale. The brains complexity is still unknown, but it is orders of magnitude larger than this.

[1] The final product is obviously not simple, due to the ways in which these simpler parts can be combined and configured

Wieke:
What a coincidence, I just received a study book on machine learning. (I am studying AI.)

rolfwesselius:
So is it sentient or did it just do as it was programmed to do?
No jokes please this is a serious question.

It just did as it was programmed to do. It's an advanced datamining algorithm.

Funny thing is, they checked it's performance using images off imagenet and had an accuracy of 15%. But it's accuracy for human faces and kitties was much higher because they trained it using video from youtube. (Basically it's a roundabout way of saying youtube has lots of cats.)

Just so nobody is confused: that 15.8% accuracy is on an evaluation set with 20000 classes (so a blind person would get roughly 0.005% correct), and it was an improvement of 70% over the previous state of the art (so apparently that got less than 10% correct).

Also, I don't think "is it sentient or did it just do as it was programmed to do?" is a good question to ask. I'm not trying to make fun of rolfwesselius for asking it, because this is a very real issue among many people, even within the field of AI itself, and my opinion is far from universally supported. However, the question presupposes that something that just does what it was programmed to do cannot be sentient. If that is the case, then no computer program could ever be sentient, because every computer program does exactly what it was programmed to do (which might not the same as what the programmer intended, but that is a different issue).

Of course, many people believe that machines can never be sentient. But then you would need to answer why they couldn't be. What is it about humans that makes us sentient that is fundamentally impossible to recreate in a machine? I cannot think of anything. Can you?

But can it invent the concept of love?

Capcha: Easy as cake
<.< >.>

Jordi:
Of course, many people believe that machines can never be sentient. But then you would need to answer why they couldn't be. What is it about humans that makes us sentient that is fundamentally impossible to recreate in a machine? I cannot think of anything. Can you?

Nope, I don't think anyone that ascribed to a purely naturalistic world view could. And of course the question is a bit silly, we have no good definition of sentient. Even if we had a computer would not be able to defy it's programming much in the same way we are not able to defy the biological/neurological/psychological laws that govern our behavior.

(Finally a captcha I can figure out in one go. Heh, can't wait for these things to become utterly useless.)

Old Long Johnson!

Old Long Johnson!

Wieke:

Jordi:
Of course, many people believe that machines can never be sentient. But then you would need to answer why they couldn't be. What is it about humans that makes us sentient that is fundamentally impossible to recreate in a machine? I cannot think of anything. Can you?

Nope, I don't think anyone that ascribed to a purely naturalistic world view could. And of course the question is a bit silly, we have no good definition of sentient. Even if we had a computer would not be able to defy it's programming much in the same way we are not able to defy the biological/neurological/psychological laws that govern our behavior.

(Finally a captcha I can figure out in one go. Heh, can't wait for these things to become utterly useless.)

The simplest marker of sentience, I think, is the Turing Test, where basically you put an AI in a chat server, and have it actively engage with people and give no indication at all that it's not a real person. Once we can do that, who knows. The computers win at that point.

I wonder if will identify a gun in the 10 milion youtube pictures it was offered.

More importantly, what if the google team intentionally removed guns from the 10 million videos available to the machine?

I wonder if will identify a gun in the 10 milion youtube pictures it was offered.

More importantly, what if the google team intentionally removed guns from the 10 million videos available to the machine?

As a computer science student I just had a small geekgasm.

Sixcess:

It's also fair to say that bombarding someonething with ten million things on Youtube over a mere three days is probably an excellent way to convince it that we should be wiped out.

Good going, Science. Hopefully my kitty will speak up for me when the robot revolution comes knocking at my door.

Clearly the computer recognised who is really in charge in the world.

And they say mew.

I, for one, welcome our fuzzy, hilariously-grammar-challenged tiny robot overlords.

 

Reply to Thread

Log in or Register to Comment
Have an account? Login below:
With Facebook:Login With Facebook
or
Username:  
Password:  
  
Not registered? To sign up for an account with The Escapist:
Register With Facebook
Register With Facebook
or
Register for a free account here