The Next Big Thing

[sje]

The Next Big Thing

The next big thing for competitive applied AI programming is not go or poker or some other traditional human game. No, the next big thing is building a programming toolset for something that has a future that's rather more practical; a toolset that can be used to implement the software side of a humanoid robot.

Now while we're not quite at the point where the mechanical hardware is available for experimentation, we can with reasonable confidence simulate such hardware. And with that, we can start on the programming.

Consider a software simulation of humanoid robot that's equipped with simulated hardware: video cameras for vision, microphones for hearing, various motors for powering limbs, tactile sensors, accelerometers, and whatever else might be needed. This robot simulation would have a good formal specification and an open source implementation. The simulation itself would be embedded in a standardized world simulator that would provide sensory input and allow for feedback.

Instead of programming chess engines and getting an ever diminished return, coders could program a standard robot simulation to handle competitive tasks. The simplest task would be something like walking from one corner to another in a (simulated) room where speed would be the competitive metric. A much more advanced task would be to program the robot simulation to perform as, say, a wrestler for combat with another simulated robot. And instead of chess, we could have two teams of simulated robots to compete in a football game.

[towforce]

The next big thing for competitive applied AI programming is not go or poker or some other traditional human game. No, the next big thing is building a programming toolset for something that has a future that's rather more practical; a toolset that can be used to implement the software side of a humanoid robot.

Sorry to be argumentative, but I don't think it will happen that way.

Each time a robot becomes better at doing general tasks (e.g. housework), people will buy it even if it is expensive. Improving AI (e.g. interpretation of live images from a camera) will make robots better at general tasks. However, I think that it is unlikely that early models will be humanoid in form.

[sje]

Each time a robot becomes better at doing general tasks (e.g. housework), people will buy it even if it is expensive. Improving AI (e.g. interpretation of live images from a camera) will make robots better at general tasks. However, I think that it is unlikely that early models will be humanoid in form.

http://en.wikipedia.org/wiki/Asimo

[towforce]

Each time a robot becomes better at doing general tasks (e.g. housework), people will buy it even if it is expensive. Improving AI (e.g. interpretation of live images from a camera) will make robots better at general tasks. However, I think that it is unlikely that early models will be humanoid in form.

http://en.wikipedia.org/wiki/Asimo

Such frivolous items will, like AIBO, become abandonware IMO, whereas machines that can do useful work will continue to be improved.

[MattieShoes]

Robocup has been going on since at least 2002, and they've done stuff with humanoid robots, non-humanoid, swarm, central intelligence, different sizes, etc.
http://www.youtube.com/watch?v=FOpvM9JmRNY

I watched some stuff on it, and robots on wheels are much more fun to watch.

And it's fun to imagine some sort of future of robotics like Asimov's "I, Robot", but seriously, This is the past, present, and future of robotics.

[mschribr]

I don’t think an android is useful unless you want something like SVEDKA vodka Robot Girl.
I think the next useful AI project is computer translation. The only problem I see is how to decide who is a better translator. It is not as simple as seeing who wins a game. It is subjective call. But a competent computer translator would be a useful thing to have.
Mark

[Bill Rogers]

A simulation? A waste of time. I started building a robot over 7 years ago.
It started with a small kit from a company called Evolution.com. Basically it was only a wheeled platform designed to hold a laptop. It came with the software that enabled it to be programmed to do hundreds of tasks and it could also understand English commands in a small way. It also had one of the best facial recognition programs available to the public at that time.
At first I would not have purchaed the thing but then I found another piece of software called Ultra Hal Assistant. This was basically a chatter box program except that it learned from every thing that you said or taught it. I discovered that there was no problems running both programs at the same time so now I had a robot that could be progrmmed to do what ever I wanted but one that I could talk to at the same time.
It has been fun and challenging at the same time but if you do a little research you will find that there are hundreds of people like me all over the world doing exactly the same thing.
Simulations, no, the real thing, I've got one already.
Bill

[mschribr]

A simulation? A waste of time. I started building a robot over 7 years ago.
It started with a small kit from a company called Evolution.com. Basically it was only a wheeled platform designed to hold a laptop. It came with the software that enabled it to be programmed to do hundreds of tasks and it could also understand English commands in a small way.

Does the robot have hands? Can you tell the robot to take the dishes off the table and put them in the dishwasher? Can you tell it to take out the garbage? Or can you tell the robot to put the laundry in the washing machine?
Mark

[Kirk]

Sounds like Skynet to me