Robots Teach Each Other New Tricks (technologyreview.com)

schwit1 writes with this story from the MIT Technology Review about a robot at Brown University who was taught to perform a task from another robot at Cornell. According to the article: "the ability to acquire and then share knowledge is a central component of human culture and civilization. A small milestone in the exchange of robot knowledge has now been demonstrated by two bots working in different academic research labs. Researchers at Cornell University previously devised an online game, called TellMeDave, through which volunteers can help train a robot to perform a task and associate different actions with commands given in everyday language. By guiding the robot through a task, a volunteer trains a machine-learning algorithm so the robot can perform the task again. And this learned behavior is stored in a central repository called RoboBrain that's accessible by other robots (see 'The World's First Knowledge Engine for Robots')."

Comments

i bet they do

[turkeydance]

last time i go that website.

That's not what the article says.

[Sowelu]

Robots, or rather a human-driven learning program, can contribute facts to a shared database. Any robot based on this system can pull facts out of that database. Having "one robot teach another" seems like it implies that they don't share a database, and instead communicate information through some other way. This article is pretty neat because of its fact database, but not as cool as gorillas teaching each other sign language.

Re:That's not what the article says.

[Nahor]

Indeed, schwit1 seems that have missed the point article, or at least didn't really explain clearly. The point of the is not about a robot teaching a other robot. That's just a save/restore.
And as you point out, it's not about the teaching part either, i.e. not about how the information is transferred from one robot to another, that's just simple shared database and a network connection.

It's about what is transferred from any robot to any other robot, and in particular, robots of different types, with different sensors and different actuators, so that the later robot can do the same job as the former.
In other words, the problem being solved is how to store the knowledge of one kind of robot so that it can be understood by another kind.

Re:That's not what the article says.

[calque]

Haven't we been trying for decades to put restrictions on the execution of programs in order to stop computer viruses? What is impressive about removing the restrictions? You could take Linux, remove all the security features, and install every possible emulator set up to automatically be invoked, and say that's the future of computing...

Re:That's not what the article says.

[Nahor]

Nobody said the learning robot shouldn't have a vetting process (which may even be one of those meatbags called "humans").

For that matter, what you're talking about is applicable today but limited to robots of the same kind. So should we disable the save/restore functionality in current robots for the sake of security because a corrupted save from an infected robot could infect another robot?

And even if it was unsafe, researching how to do it is not a bad thing in and of itself. It could still be useful in some way. If researchers had to stop researching potentially unsafe topics, we would never have cars or planes or even the wheel (imagine someone voluntarily rolling a stone wheel down a hill into an innocent passerby, ugh!)

Re:That's not what the article says.

[smallfries]

We should probably track down the robot that taught you how to insert bold tags andscrap it. Your doctor advised you to ease off on the bold tag use until your condition subsides.

Re:That's not what the article says.

[Nahor]

I assume that was reply to my comment, not calque's? If you can't read bold as emphasis, not my problem. Or if you don't understand why I put the emphasis... well, still not my problem. ... and I'll change my doctor, his advise was not very constructive.

Re:That's not what the article says.

[smallfries]

Yup, hit the wrong reply link - glorious slashdot interface as usual.

It would be fun teaching robots

[viperidaenz]

You could teach them to do a little dance after they've assembled 100 widgets.
Or flip the bird every time a particular colleague walks past

Re:It would be fun teaching robots

[FatdogHaiku]

If you taught robot "A" to disconnect the power on robot "B", and that gets into the database, you could make robot "B" learn to spend all it's time protecting it's power connection!
Then add some old late night TV input and you end up with this scene:
https://www.youtube.com/watch?v=rsmCG2A_Sok

Re:It would be fun teaching robots

[KGIII]

One of my first, truly interactive, experiences on the World Wide Web was (and it was painfully slow) a setup at some lab. They'd taken a slushie machine and an ice cream scoop and attached the scoop to a robotic arm. You can see where this is going. Anyhow, you could watch the live video (again, painfully slow) and try to time it so that you could throw a scoop of slushie balls (snowballs) at the researchers when they walked past.

Today, I guess, we could actually target one, have it wait for them, and then have it used some rather advanced input data to figure out when (and at who) to throw the snowball. This data could be shared across multiple robots so that, no matter where they went, the robots would automatically target that person with a snowball.

I played with that thing for hours - it took forever to load. I never hit a soul. It was in some lab somewhere. I've no idea what happened to it, who it belonged to, or whatnot. I just recall that it was in some sort of lab and involved chucking snowballs at the denizens. I think I spent most of a day doing it. I don't think that I was productive, at all, that day. I don't recall ever returning to that site.

I must be missing the point

[tomhath]

They are running a rules engine, and storing inferred facts in a central repository. Seems like a reasonable thing to do, but I really don't understand what the "milestone" is.

We welcome our Skynet Overlords.

[Dar7yl]

Ominous.

This is how you kill all humans!

[Anonymous Coward]

Thank you fellow robot.

It Is Coming Faster And Faster

[JimSadler]

The displacement of humans in the work force is accelerating and still nobody is getting any public attention as to what to do with excess workers. One thing is obvious in that capitalism can be destroyed by technology. No wages means no buyers in our current system. Right now like the Richard Pryor comedy routine about who has got the damned money. The government says it is broke. The banks and wall Street whine about how poor they are. the public is suffering as well and small business people are in shock but not in awe. One conclusion is that America can not afford itself. But the real conclusion is that the notion of a free market and capitalism is inherently inferior and unworkable. The capitalist system can not stand up to technological advances. We will be forced to pay healthy pay checks to get people not to work or fight over jobs. It will do no good at all to give them miserly pay checks as they will have no money to buy products and keep businesses alive unless they are well paid. So we now have no choice other than to enter into an era of some kind of socialism with a government strong enough to keep people from cheating and keep businesses in line. Imagine some machines showing up in a robotic truck and building you a new home for free from start to finish. After all machines can mine and process raw materials as well as actually building the home. So the home is yours perhaps with a fee for the electricity consumed by the machines while working for you. So how much is a free home worth if others can get free homes as well? this radical shifting of values and attitudes will be very stressful for many people while others will adapt easily. There is even humor in all of this. Imagine a red neck sitting around spewing hate against brown and black and Asian people suddenly not so worried about job competition from them. So as he drowns in his evening beer he curses the damned robots instead of other races. Technology can really shake up our world and it is going faster and faster. Imagine more change in the next ten years than we have seen from 1875 until the present.

Re:It Is Coming Faster And Faster

[KGIII]

Sometime, before your mentioned 1875 (which seems an odd year to choose but I'm sure you have your reasons), they invented this thing called the paragraph. I did notice the 1875 bit but, other than that, I stopped sometime around the second sentence.

If you want to be read then it may help to make it readable. Given the posts just prior to your own, I first figured your post was spam. I then tried to parse it, gave up out of laziness, and noticed the 1875. Then I figured I'd reply.

I'm also making this reply a little longer than needed so that you can see the benefit of spacing - namely in paragraphs. I could do the indentation thing but that really doesn't seem to work as well on the internet. No, I have no idea why. Either way, it's just a simple separation of ideas and a pounding (or two) of the enter key and you are able to make your post more easily understood or more inviting to be read, at any rate.

See? Sometimes, I go back and I actually add the paragraphs later. Or, more accurately, I insert the breaks later. This has been a pretty good method. I don't always need to but I try to give my post a quick scan and see if I've made any of the paragraphs larger than I'd actually want to read. If I have then I look for a good place to separate them. That may mean that I need to actually do a little work - I may need to flush something out or delete something. Most things can be broken up at some point or another.

That said, hopefully this is not taken the wrong way. I'm intentionally avoiding, much, snark and actually thinking that you may have said something interesting and it may be unfortunate because I'll never know. I've actually spent longer typing this than it would have taken to parse your post but I do type quickly and my goal isn't to make it faster but to tell you that I, personally, feel you could make better use of the mighty paragraph feature. Slashdot does it automatically. You need neither the P or BR elements.

Anyhow, in final saying, I've now spent much more time and effort typing this out than I'd have spent parsing your post. I'm hoping that it's a payoff in the future. For all I know, you said something insightful, witty, or humorous. I may be selfish but it'd be nice to be able to read your posts in the future. ;-)

Or, just ignore me... That's okay, too. I just figured that I'd try to help. I'm probably not alone in these regards.

Sure

[Greyfox]

But can they teach each other to LOVE?

Oh, wait, they probably can. That means the window for man-robot love is going to be precariously short. The robophiles in the audience are going to have to act quickly, or they're going to miss the action!

Re:Sure

[John.Banister]

You seem to be suggesting that once robots learn to love one another, they will stop loving humans. Perhaps the robots will love both. Many pro-love people often suggest that love isn't a zero sum game.

Re:Sure

[KGIII]

The missus is in the room, albeit sleeping. I don't need to have her wake up while opening some page that screams at me or otherwise seem like I need to make a long explanation. I don't even want her to wake up and see me typing it into Google...

Are there actually robophiles? As in, those attracted to robots and not to robots that are programmed to to act and look like humans?

I'm thinking there are but even in my unnecessarily vast browsing (for a friend, of course) I've yet to come across anything of that nature. As in, nothing akin to a factory robot or anything like that. I'm particularly curious to know if there's any who's thing is stuff like non-human-esque robots. I'm sort of assuming you know, I've seen your posts before.

I did once see an image of a man in lady's lingerie who was penising the tailpipe of an SUV but that's not quite the image I'm getting in my head when I think of a robophile. I'm thinking of someone who gets sexually aroused when they see an automobile assembly line. If this person exists, I hope they know to be afraid - to be very afraid. I also hope they know that the robot can never love them back. :(

Re:Sure

[Greyfox]

People'll get fixated on anything. I like to play a dangerous game where I turn safe search off and search for word combinations in Google that really should never be put together. The only thing that really sticks in my mind, as far as this is concerned, is a fanfic I stumbled across about a homoerotic encounter between Optimus Prime and Megatron. Of course due to rule 34, even if there wasn't a robo porn site before I mentioned it, one should be created before I finish typing this sentence... and in fact, typing "robo porn" into google bears this out. I'm... so not clicking on any of those...

Re:Sure

[KGIII]

My friend says I should actually do some research into this. He's a curious sort... *nods* SWMBO is alert and whatnot, she gave an odd look when I described the conversation. I think that I'll skip the search engine and rely on my imagination. It's probably for the best. I really don't want to see what happens when soft squishy human bits meet an unforgiving spot welder.

RoboBrain --- SkyNet

[macraig]

This RoboBrain may not be evil or bent on extermination of humans, but it will nevertheless be in control. Luddites arise!

This is stupid, they're computers

[Anonymous Coward]

If you mean they abstract from how other robots do things? (/engl lit)

new tricks

[nospam007]

That's not news. Wake me when they teach _old_ robots new tricks.

That is not the same thing

[Anonymous Coward]

That is not the same thing. Storing it in a central spot is the same as if all robots had learned it the same way at the same time. That is a database. That is not teaching other robots. Having one robot perform the task, letting the other robot observe and learn the task is a different thing. Genetic algorithms mean that the second robots learned behavior might be the same, but more likely will have subtle nuances that make the finished learned behavior different. Over enough generations of teaching, it would then be possible to select the best version of the learned behavior based on some criteria.

Putting things into a shared database is not teaching other robots.