DailyDirt: Computers Becoming More Like Us
from the urls-we-dig-up dept
The field of artificial intelligence is steadily making progress and developing software that can perform some pretty impressive tasks. Still, most AI projects aren’t quite ready to convince everyone that computers can be good at general intelligence. But to some folks, it’s only a matter of time before robots are going to take over — driving cars better than us, beating us at chess/poker/go, stealing all the manufacturing jobs. Here are just a few more examples of artificial intelligence getting smarter.
- Algorithms deep inside the secretive Google X lab have been watching millions of YouTube videos — and have learned to identify cat faces. This feat was accomplished with unsupervised learning — so no one explicitly programmed anything to look for cat faces. Maybe those computers figured out for themselves that kittens are good for productivity… [url]
- Eugene Goostman is a chatbot that has come very close to passing the Turing Test by fooling a panel of human judges into thinking it was a real boy 29% of time (Passing the Turing Test requires a 30% fooling rate). This chatbot was given the fake personality of a 13yo boy living in Odessa, Ukraine — so expect next year’s competition to include a lot of fake teenagers with strange idiosyncrasies. [url]
- DeeChee the iCub robot is learning how to talk like a human baby by listening to adults speaking and babbling until recognizable words form. DeeChee is a project in the field of embodied cognition — which asserts that cognitive processes are shaped by the bodies in which they occur. This robot doesn’t think like a human baby, but it could help understand how biological brains create language. [url]
If you’d like to read more awesome and interesting stuff, check out this unrelated (but not entirely random!) Techdirt post.
Filed Under: ai, artificial intelligence, chatbot, embodied cognition, eugene goostman, robots, turing test, unsupervised learning
Companies: google
Comments on “DailyDirt: Computers Becoming More Like Us”
Well
I for one welcome out cat face recognizing robot overlords!
Re: Well
so you would know that any robot that has gone crazy can be killed with the flash of a camera..
Re: Re: Well
“You know, you’re right. This truly was the best vacation ever. Now let us never speak of it again.”
embodied cognition
that’s an interesting concept.. so do nerds become smarter b/c they inhabit nerdy bodies?
Ray Kurtzweil anyone
It reminds me of that Our Lady Peace album that was also inspired by Ray Kurtzweil’s book called The Age of Spiritual Machines.
Osamu Tezuka
I’m surprised Osamu Tezuka hasn’t been mentioned yet, considering we might actually be taking step in creating a “robot” world. Wonder what he’d think about all this?
Er, steps is what I meant to say.
I would like to take this chance to say that the Turing test, although invented by a brilliant mind, is complete and utter garbage. Why? Because people with a categorical knowledge of a subject are routinely voted as being AI, and chatbots that communicate about nothing in particular score the highest on the human scale.
Any intelligence shown by the contestants is immediately interpreted as artificial. Therefore, this isn’t a test for artificial intelligence, this is a test for artificial shallowness and ignorance.
Re: Re:
very true, artificial intelligence should really be called artificial smartness, being smart and being intelligent are not the same things.
as you’ve pointed out, it would be worth of not if there was a computer capable of hatred, spite, fear, joy, love, curiosity, abstraction, empathy.. all things computers cannot do, but humans (and animals can).
humans can do tasks but computers can also do, like drive a car, but it is arguable that humans are using intelligence to drive, they are acting more like a computer than a human in most tasks. (like driving).
it’s the HUMAN things that humans do that computers will never ‘get’..
Re: Re:
being a test based on a converstation and a guess about what or who you are conversing with it’s not a good test, (or actually a test at all)..
you ask the computer “what does love feel like?” it is going to look up some database and provide you with a standard answer, but it will never be able to tell you what love actually feels like, because that is an emotion that requires intelligence.
Eugene Goostman is a chatbot that has come very close to passing the Turing Test by fooling a panel of human judges into thinking it was a real boy 29% of time (Passing
the Turing Test requires a 30% fooling rate).
SO if you totally ignored the chat bot, and flipped a coin, you at least would achieve a 50% guess rate..
So to pass the test you have to ‘guess’ a value (negative result) way less than 50% of the time….
Not much of a test..
If you pick things that humans can easily do, but in doing them are in fact acting like computers is not much of test.
SO computers can drive cars,, so what, humans can act like computesr enough to drive them as well.
But it’s going to be a very long time before the computer in your car feels fear or anger when it is cut off on the road..
No matter how well a computer can control a car, it is never going to decide by itself that it feels like a burger, and drive to the shop and buy one.
computers can be smart,, but not intelligent
when the cat face computer does an identity theft, and orders a kitten from craigslist, along with milk and a bowl.. and refused to give the kitten back because it loves it, then that will be interesting.
did the program call it a cat ? and comment “I like cats!”
it’s probably an ANN, particularly good at pattern recognition.
For a computer, being able to reliably identify a cat face is actually REALLY difficult, because it requires you to try and define a cat face say compaired to a dog’s face.
-Size isn’t reliable, you have chihuahua’s to Irish wolfhounds and the varying closeness of the shot from the camera throws that out the window anyways.
-Shape isn’t reliable, cats have round faces, except when they don’t which is just as common.
-Color doesn’t help.
-MAYBE the cat eyes, but that isn’t reliable as well.
Can anyone come up with a cat description that doesn’t also describe another animal?
This isn't all bad
I still believe that having robots do things for us isn’t a bad thing. I WANT robot drivers. Have you SEEN how absolutely AWFUL a HUMAN driver is? Even people who drive 300 laps per day in the exact same circle (NASCAR) frequently have wrecks that kill them! For something so readily predictable, that’s disgraceful. A robot would never make such a catastrophic error when the task is doing the exact same thing over and over and over…
Cars aside, I don’t think allowing robots to replace manufacturing jobs is bad either. I want a future like WALL-E minus the pollution. Think about it: aside from the pollution, WALL-E’s future is pretty damn Utopian.
Of course, I also want to eventually be able to upload my mind into a robot chassis and live forever, so if nothing else I want to see advanced AIs because any system capable of running a sufficiently advanced AI should be capable of emulating a human brain with the right modifications.
Seriously though, this isn’t all bad news. Some of this is awesome.
Re: This isn't all bad
I didn’t mean to suggest any of these developments are necessarily bad. Technology that produces better tools can be used in both good and bad ways, but it’s generally better to have more technology than less….
We just have to be careful about how we use the technological marvels we build.