A Piece-By-Piece Guide to the Most Advanced Bots

XopherMV cuts-and-pastes from Wired: "In an article from Wired, 'Consider the progress of just the past 15 years. There are now robots that can get around on two legs, participate in simple conversations, and manipulate objects in rudimentary ways. Of course, we don't yet have a bot that can navigate downtown Manhattan, tie its shoelaces, or even tell a chair from a desk. MIT's Cynthia Breazeal holds out hope that within five years, robots will cross a critical threshold, becoming partners rather than tools - in other words, we'll have friends, not appliances.'" Reader ptorrone adds: "In Los Angeles, CA at the Century Plaza Hotel for the 4Site conference, our favorite robot vacuum/military supplier, iRobot, showed off the tactical mobile robot! The 'Tactical mobile Robot' has its own brochure and site: www.packbot.com. The rad thing about this platform is its skateboard design, where it appears to support various plug-in modules. Here are some photos of the packbot!"

Comments

you're kidding!

[Savatte]

Of course, we don't yet have a bot that can ...tie its shoelaces

The only reason is because velcro is more efficient

Friends in Five Years?

[stinkyfingers]

Take a look at the difference between ELIZA and ALICE, for example. ALICE is still just a pattern-matching language parser, just as ELIZA was from decades ago. Both qualify as being able to partake in simple conversations. ALICE simply has more comupting power available to it - power that it wastes on XML, I might add. Is there absolutely no chance that, in 5 years, there will be a quantum leap in AI that allows us to go from ALICE to something that can carry on a meaningful conversation? I won't say that, but it won't be more meaningful than give commands.

Hardly qualifies as "friends, not appliances". In plus, if a robot ever figured out that it was smarter, stronger, and better looking than me, it would turn around and kick my ass.

Yeah, OK LADY

[stratjakt]

Cynthia Breazeal holds out hope that within five years, robots will cross a critical threshold, becoming partners rather than tools - in other words, we'll have friends, not appliances.'

There's been a Cynthia Beazreahal, or counterpart thereof, saying this since the 50s.

You all hold out for your robot friends, but it's a friday night and I plan to go out drinking with some live human ones.

5 years?

[Rorschach1]

I think this is one of those things that's going to stay '5 years away' for the next 30 or so.

Algorithmic functions like balance have improved, sure. But how much real progress have we seen in fields like speach recognition and machine vision? Just look at the results of the DARPA Grand Challenge. Or my stupid cellphone with its voice dialing. It's only got half a dozen samples to compare against, and yet it takes about three seconds and never manages to distinguish between 'Keri' and 'Debbie', and won't ever accept 'Lee' (or any other one-syllable names, for that matter) at all.

It was true 30 years ago, and it's true today. AI is bogus.

The only branch of AI that I have any faith in is neural networks. We've got pretty good evidence that they WILL work if we figure out how, but I don't see that we've gotten much closer to that point in the last 30 years either.

As for working with machines as partners, STOP TRYING TO MAKE MY TOOLS SMART! They're tools. Make them do what I tell them to do, not what they THINK I'm trying to do. Hell, working with dogs is a challenge sometimes, and they're orders of magnitude smarter than any software that's out there now.

Re:5 years?

[starm_]

I have studied neural networks extensively and believe me these do not have the potential to revolutionize anything.
NN are as simplistic & bogus as the next thing. Other methods like Support Vector Machine has shown to be more powerfull. Not to say that there isn't room for improvment or that AI will nerver be fruitfull. Its comming, slowly but surely. here are a few reference to interesting AI research:

1
2
3
4

Re:Friends not appliances?

[BK425]

Not to go all serious on a promisingly humorous start... but perhaps those of us without friends could deal with those issues (you know who you are) as seperate from technology design issues. Because they are.
"we'll have friends, not appliances." is a _seriously_ bad goal. I -want- an appliance that I can order to clean out the hold of an oil tanker. I do not want to order a sentient being to do unsafe or tedious and boring things. We have plenty of sentient beings, and they enjoy reproducing fairly efficiently. It seems really obvious that applying technology to create sentient, or even sentient like, life is a bad thing.

Re:Good robot.

[kfg]

Robots will become "friends" instead of "tools" the day the first one says, "No, I will not help you move", or "Not tonight, I have a headache."

The very next day that robot will be sitting turned off and in the closet, or back at the shop to be "repaired."

I'd guess there are maybe 3 people in the world who really want a robot "friend", and they're both socially awkward roboticists.

"I am so happy I am standing beside myself."

The rest of us want Johnny 5 to vacuum the floor, do the dishes, pick up the laundry, cook dinner and shut the hell up when we tell him to.

KFG

Self Aware

[Usquebaugh]

Until we can breed an AI that is self aware robots will continue to be the sum of their programming. Nothing wrong with that but it's hardly anything new. all that's happening is that hardware is getting better.

Re:Self Aware

[protohiro1]

There are plenty of animals that are capable of solving a lot of the "hard" AI problems without self awareness. My cat is not self aware, but she can easily jump on to small things, recognize items by there appearence or smell, catch moving objects, distinguish between food and non-food and she can figure out how to use a cat door. More still, she can understand that when I put a treat on a shelf it remains on that shelf when I leave and then attempt to climb up to get it. If I put something she wants behind a door she can see that there is a problem and then attempt to solve it.

what are they smoking?

[hkomsuog]

I, personally, cannot stand people from MIT keep saying things like "robots will do this and that in so many years" Rodney Brooks (the current Director of the MIT Computer Science and Artificial Intelligence Laboratory and Chairman and Chief Technical Officer of iRobot Corp) came up with his infamous "subsumption architecture" in late 1980s and claim at that time that it was solution to legged locomotion control. Though, as far as I know he and his group has failed to show anything more than several slow and limited robotic implementations in the last 2 decades. This and similar approaches that claim to design controllers based on primitives (CNNs or Area, et al., or BMPs of Kirchner et al) all lack analytic framework. Hence ,the design process has a big hole in the middle which needs to be filled up by the intuition of the designer. The resulting controllers tend to be very complex and offer no basic understanding. So, I find it rather comical to hear them keep saying "robots will roam the world in so many years." We are barely scratching the issues. As far as I know the only thing MIT offers these days seems to be a robot that demonstrates some facial expressions(Cynthia Breazeal's Kismet). Big deal. [I know they are doing other things like COG but that project doesn't even address the locomotion issue] Without legs it wont be happy anyways. There are even some MIT people who critize these projects as waste of time. If anybody it is Mark Raibert of MIT leg lab who made a siginificant contribution to legged locomotion back in early 1980s. I don't remeber him going around in publicity rounds and say robots will conquer the world. Such ungrounded comments can ultimately hurt the field. People are already quite edgy when it comes to technology. Anyways, just my 2 cents...

Re:what are they smoking?

[Animats]

The previous poster has some good points.

Brooks did some great insect-level AI. It's purely reactive, with no world model. That was good work, and a reasonable reaction to the logic-based planning crowd then running MIT AI. But then Brooks started going around saying that the reactive model was far more powerful, and started making human-level AI noises. From this came Cog.

When Brooks came by Stanford to talk about his plans for Cog, I asked him "Why don't you go for mouse-level AI", something that didn't seem totally out of reach. He said "Because I don't want to go down in history as the man who created the world's greatest robot mouse". As one of the grad students on the Cog project said, "It just sits there. That's what it does. That's all it does". And, years later, it still doesn't do very much.

The model-free approach is just too dumb. With no world model, you can't get beyond insect-level AI. That approach works mostly for creatures in environments where inertia doesn't matter much. For insects, banging a feeler into something is fine. For large animals, you get bruises or worse. As creatures get bigger, faster, and stronger, they need models with some predictive power, so they can avoid mistakes before damaging themselves. I tell people in academia that you need to be "less formal than Latoumbe (who formerly headed Stanford's robotics operation) and more formal than Brooks". The game development community has absorbed this lesson, but it's only starting to get through to the robotics community.

Raibert's work on legged locomotion was very impressive. I'm very familiar with that work; I've done some improvements on it. Raibert had one great insight - balance is more important than gait. People have been studying locomotion for a century, and almost all the studies center on gait. Raibert realized that balance was more important, and built a one-legged hopper to force the issue.

But, in fact, the way Raibert does locomotion is very simple. There are two controllers, both simple hand-tuned PID loops, and a state machine that swiches between them. This can handle simple locomotion on the flat, and some preprogrammed moves like flips, but it doesn't generalize. I'd expected the adaptive control people to pick up from where Raibert left off, but so far, nobody has really done that.

My insight there was that slip control is more important than balance. On the flat, traction control isn't a big deal, but on hills or rough terrain, traction control dominates balance control. That's what legs are really for. If you add automatic traction control to Raibert's approach, legged running on hills becomes possible. Otherwise, you slip out climbing hills.

Raibert himself left MIT and did a startup company, Boston Dynamics. But they ended up selling products to DoD which are game-like kinematic simulations. They don't seem to stress dynamics work any more.

The MIT Leg Lab was taken over by Gill Pratt, who was more of an actuator and controls guy. He didn't accomplish much. The next head of the Leg Lab was some guy who was into prosthetics. The Leg Lab now seems to be defunct. Their web site hasn't been updated since 1999.

Re:Friends not appliances?

[Bingo Foo]

No kidding. Too many people today already think of friends and neighbors as being disposable or at least interchangeable in some sense. Imagine if technology deliberately blurs the line between people and tools.

iRobot?

[Felinoid]

Something frightening when iRobot starts violating the 3 laws of robotics before it's even built.
(Military application would violate "Cause harm" and "alow harm by inaction")
(Not exact quotes of course I'm being lazy)

Robot friend? So I finnally get to have a happy chearful elevator that thanks me every time I enter it? Or better yet a paranoid android.

Re:Good robot.

[kfg]

Oh, Hi! Where's the other guy? (the "3" was a typo)

Sorry, I couldn't resist the setup.

How many people in the world have a very strong attachment to their pet?

Myself for one. I don't forsee a robot replacing her. I also have a strong attachment to my guitar. I might even speak of it colloquially as my "friend," but it's just a tool. I'll be heartbroken when it "dies," it's been my "friend" for more than 20 years now.

I'm not sure you get the point I'm driving at. I'm not speaking of the odd attachments that people get for various things, even inanimate things.

I'm speaking of the way things are treated because "friend" is about behavior, not simply attachment.

I went out without my cloak last night, even though it was chilly enough to be uncomfortable without it.

Phoebe was sleeping on it.

My Fluffy 3000 would have been ordered off or manhandled off without a qualm. It's a machine. It won't scratch my furniture either, because I won't allow it too. I'll program it to behave as I wish. If it does not behave as I with it will be considered "broken."

My Barberella 3000XL Platinum Blonde Edition with Turbo Boost and Bluing for extra Whiteness will never get a headache, not because she couldn't be programed to, but because I wouldn't accept that programming. She will do the dishes when I ask her to, cheerfully, as my "friend" every time I ask her to.

That's not a friend. That's a tool.

Sure, I'll be, ummmmmmmm, "attached" to her, who wouldn't be? I'll probably even call her, and even think of her, as my "friend" to some extent.

It's a very warm and fuzzy feeling illusion, isn't it?

KFG

Re:Friends not appliances?

[Smidge204]

Two flaws I see in your comment right away... sentience does not automatically imply that is it capable of viewing any task as boring or tedious. many humans do what you probably consider boring and tedious for a living (like assembly line workers, digging ditches, etc) and sometimes actually enjoy it. All you need is a good attention span and the ability to focus on a task, and it is no longer boring or tedious.

Second problem is safety. A machine, assuming it has been properly designed for it, is at a MUCH lower risk of damage for a given task than a human. Cleaning out an oil tanker hold is a perfect example, and so is changing out nuclear reactor cores or repairing vehicles is space.

The added advantage of a sentient machine is that the "mind" can be seperated from the "body" if you are really that concerned about it "dying" during a dangerous task, combined with the advantages of being a machine in the first place as given in the above examples. You can always build it a new body, which is a bit dfficult to do for meat and bones.

Does this mean I'd want to discuss the morning headlines with my toaster? No, not really, but poo-pooing the development of sentient machines as a whole is a big overboard.

(And on a personal note, yes we have plenty of sentient, STUPID beings on this planet who essentially do nothing BUT reproduce efficiently. So those qualities are not always a good thing IMHO)
=Smidge=

Re:Nothing new...

[handsome devil]

Another threshold found in robot development is when robots become *too* close to looking human.

Up to a point, the more humanlike a robot looks, the more we identify with it. There is a point where the robot looks so human, people are disturbed by it. The robot looks human but lacks the spark of life.

RealDolls remind me of corpses. Oh well, whatever floats you boat...

Neural networks?

[mikelang]

I do not really believe in neural networks as they are.

By now an "artificial neural network" is to brain, as "hello world" program to an application development platform with os included.

And when you reach proper level of complexity they just become harder to build and understand (not that we always known how they REALLY work).

So please: keep with tools that we can still understand - they are EASIER TO USE!