Ten years ago, I built an ostensibly cutting-edge mobile robot from a kit. It cost close to a thousand dollars, required an attached laptop computer to operate it, and it never really worked. Sure, it could zip back and forth and make noises, but the simple task I set for it – recognizing a can of soda, picking the can up with its gripper and then bringing it to me – was an exasperating near miss. It could recognize the can every now and then, maybe head in its general direction. Grabbing: nope, unless I put the can right inside the gripper. And all the software for controlling the thing was firmly proprietary, so no hope of making any tweaks the manufacturer hadn’t anticipated. (Why yes, they did stop supporting it a few years later. Thanks for asking.)
Today, I could build that same robot for well under $100, including the computer to control it. It would be about the same size as a one of those nice solid books about Python or Ruby or Drupal, instead of bigger than a milk crate. It would run open-source computer vision software. If I wanted to save a few dollars, instead of buying a servo-operated gripper from one of a dozen or more robot-part sites, I could just 3D-print or laser-cut my own at the nearest makerspace. I could mount a sensor right on the gripper to measure the distance to the can to a fraction of a centimeter, and it would cost about five bucks. And then it could tweet to the entire universe: “Hey, I just brought Paul a soda can.”
No, I don’t have any pressing need for a bot to bring me soda cans. But I might do it just because now a rank amateur like me can. After 25 years of cool stuff being just around the corner, maybe available technology is finally catching up.