Learning Robot is "A Lot Like a Puppy"

I'm excited to report that I just read the first news about real AI that's ever made me just a little bit nervous.

First off: apparently, when artificial neural networks are used in AI, they have to be carefully limited to keep robots from doing dumb things. I'm pretty sure that at this level this just means controlling the number of completely irrational solutions the robot comes up with for a problem on the level of "There's a wall in my way," but looking ahead to the future, I like to think that this could mean one missing line of code could stand between peaceful coexistance and robots eradicating humanity.

On a more serious note, the point of the article is that through cleverly mixing artificial neural networks with the traditional "pre-programmed" approach to AI, they found something that works a lot better than either one works on its own.

Here's a quote, because this is just that awesome:

Working in the EU-funded COSPAL project, Felsberg’s team found that using the two technologies together solves many of those issues. In what the researchers believe to be the most advanced example of such a system developed anywhere in the world, they used ANN to handle the low-level functions based on the visual input their robots received and then employed classical AI on top of that in a supervisory function.

“In this way, we found it was possible for the robots to explore the world around them through direct interaction, create ways to act in it and then control their actions in accordance. This combines the advantages of classical AI, which is superior when it comes to functions akin to human rationality, and the advantages of ANN, which is superior at performing tasks for which humans would use their subconscious, things like basic motor skills and low-level cognitive tasks,” notes Felsberg.

What's really interesting is how they handle the problem of setting a robot free to learn on its own. Essentially the robot has no innate criteria for making decisions, so how does it know when it's done something "right?" Well, we tell it so. A human operator has a device with two buttons.

The good boy, button, and the bad boy button.

Wow. Can you imagine when this becomes a toy, or an ethical issue?

There's actually a video game that functions similarly. I guess it'll be harder without a magical fairy.

It's Time to Take Your Medicine.... (Robots part 2)

I saw Honda's ASIMO robot today at the Robotopia Rising exhibit at the Kennedy Center, and I have to say that the presentation was not really all it could have been. ASIMO has a handful of showcase-abilities, all of them made most significant by the fact that ASIMO is life-sized, about as big as a twelve year old human.

So keeping that in mind, the biggest challenge in ASIMO's design has obviously been balance. ASIMO can't just casually throw its weight around. Walking, running, climbing stairs, kicking a ball, and "dancing..." these are all things ASIMO can do... but slowly. And carefully.

So, ASIMO was presented in a 15 minute show that was clearly aimed at kids. Which means they had a lady using her best kindergarten teacher voice to try to get everybody really, really excited about whatever ASIMO was about to do. This is kind of a subtle robot, though. For example, when it runs, it doesn't really look like it's running. It's more of a half-comical sneaking motion.

When you build something up like it's going to be awesome, knowing full well that it isn't all that visually impressive, that makes the event itself kind of a disappointment. It was like that all throughout the show, especially when it came time for ASIMO to dance.

Now, ASIMO can't actually move its lower body very well, so what it did was slowly and randomly move its arms around until the music stopped. It was really kind of sad. And then, as the show moved into the possible practical applications of ASIMO as an aide to the elderly, the show took on the "forced" and "creepy" aspects that are almost inevitable to any discussion of using the clumsy, slow, and awkward robots we have right now with vulnerable people in need of reliable help.

What would have made the ASIMO presentation really engaging, in my opinion, was more talk about the challenges of getting it to do the things it can do, and the ways its developers failed along the way, as a means of explaining how they learned from their failures to make it work. That way, the audience would better appreciate why ASIMO isn't quite ready to shake its booty yet.

So, discussion question. Is there a difference between making a science exhibit accessible and inviting to kids, and targeting it at them? Grownups made ASIMO, shouldn't there be a layer to its presentation intended for them as well? Or is the only purpose of an exhibit like this one to get the young ones excited about technology and learning?

Lucky for me, inventor/artist-troupe Maywa Denki were also performing at the Kennedy Center, providing fun and song with their "Original Instruments" consisting of guitars that play themselves, singing "robots" powered by bellows, a kind of electric castanet-set mounted on a pair of wings, and a number of other percussion instruments based off of a basic "knocker" device that goes "bonk" and "clunk" at different pitches. The group invents and creates their own instruments (among other things), then presents them as "products" as part of their performance art motif, which hearkens back to the small electric stores of a bygone era.

Not only are the instruments innovative and funny, but some of their songs were pretty catchy, too.

Others, just weird. But even still, somehow wonderful.

I Heart Robots

On Saturday I'm hoping to go see some robots with my friend Chris, and so I wanted to get into the robot spirit. Since today is Valentine's day, I thought I would introduce you to 5 robots that I love.



This is QRIO. Right now he and his other QRIO friends are just dancing (it's all preprogrammed--not very robotic at all), but apparently he can do much much more. He can throw things, identify objects and faces, move his little fingers, protect his face from danger, right himself when he's about to fall down, and stack blocks. Intended as a little robot entertainer, the QRIO project was tragically dumped by Sony over a year ago, but rumor has it that Toyota might be picking it back up.





This is ACM-R5, an amphibious snake robot created by the Hirose Fukushima Robotics Labs in Japan. Watching it move is just amazing. The different segments are all semi-autonomous, and have a system of communicating with one another that lets them count how far down they are from the head so that they can be replaced without needing reprogramming.





This is the Toyota Partner Robot, one of the robots at the Robotopia Rising show. It plays the freaking trumpet. Having spent 4 or so years of fruitlessly trying to learn the oboe, I'm blown away by this. I wonder how it would stack up on a chair test?





This is the Cornell Healing Robot, a robot that can teach itself to walk as well as develop new ways of walking to adapt to injuries. I once bought a little cockroach robot for ten dollars that responded to any and all stimuli with "turn 45 degrees to the right," and I remember being frustrated with how stupid my robot was as I watched it wiggle in endless circles next to a simple obstacle.

This robot has no preconceived notion of how its parts are put together, and so it tries out randomly generated models of what it might look like until it gets the results it expects. According to Chronicle Online, the robot never really settles on one model for forever and always. Instead it allows all the models it tried recently to continuously compete for the role of which one best describes the robot's past experiences. Sound suspiciously like the way we think?





This is Domo. Designed to help the elderly with chores, Domo is another adapting robot. It uses a couple of different strategies to figure out the size and shape of objects so it can put them on shelves, and it also responds to human commands. Probably the best thing about Domo is the pseudo-sentient naive eye contact it makes with its human master before carrying out a task.

I've seen a video of Domo trying to make a drink, too, but considering the robot identifies the bottle by shaking it upside down, never unscrews the cap, and doesn't actually make a drink, I'm not really sure if that's a one of its more reliable features. Domo isn't really intended to ever hit the market, but the research that went into creating it should contribute to making better helper robots for our glorious, creepy robot-assisted future.