The Robots are Coming

[Grendizer]

One of the inventors of the Roomba has developed an industrial robot that is simple to use and program, and is only $2.5K. It's billed as a "game changer," although so was the Segway. Here's the article by Mashable, which contains a video.

Rise of the Robots

[Svengli]

Hmm,

In my reading and research, I've come to the conclusion that the lack of AI-type software is really what's standing in the way of even fairly small steps in robotics.

Basically, as things stand, the more complex a task a given robot is programmed for, the more programmer-hours have to be spent on the task. But it gets worse. Program a robot for two tasks, even two relatively similar tasks and you can't combine the tasks in any simple fashion that doesn't even more programming hours than the previous tasks.

I can believe their robot is "easy to program" in the sense that you can record a series of actions and have it replay them. But once you get past that stage, going further starts to be much, much messier.

[Grendizer]

Supposedly that's not how it works. It doesn't just record actions, but understands the connection between actions and outcomes, such that if you tell it to pick up cups and stack them, it knows to do that even if you move the cups around. It can see the cups and figure out what to do with them in various chaotic contexts.

Personally, I believe the chief limiting factor in robotics is energy, not programming. Both programming and the ability to implement that programming are ramping up exponentially, but it won't make much difference if effective androids require tethering to a wall plug hooked to a coal plant.