Robotic Nation Evidence

Desert challenge too tough for robot racers

Robot race ends without a winner

From the article:

A $1 million race across the Mojave Desert by driverless robots ended Saturday after all 15 entries either broke down or withdrew, a race official said.

Two of the entries covered about seven miles of the roughly 150-mile course while eight failed to make it to the one-mile mark. Others crashed seconds after starting.

Lots of people wrote to send me links like these, and the gist of the message was, "See, humans have nothing to fear from robots taking over the workplace -- robots will never be able to drive trucks."

These folks are missing the key feature of the Robotic Nation: incremental improvement.

The first article makes the point this way:

However, the distance travelled by Sandstorm and its 40 mph (65 km/h) speed still surpassed the achievements of any previous autonomous vehicle.... The Department of Defense had to date come up with an autonomous truck that travels at 12 mph (20 km/h) for a maximum of six miles (10 kilometres).

As of March 13, robots can now drive 3 times faster and one mile further than they ever have before.

DARPA plans to rerun the Grand Challenge in 18 months, and it plans to double the prize to $2 million. Think about what will happen in those 18 months:

• Because of Moore's Law, the computers used in these robotic vehicles will have twice as much computing power, twice as much memory and twice as much disk space available.
• The research will advance because of the lessons learned in this race, software will get better, etc.
• Many more teams will be competing because of all of the publicity this race got. For example, dozens of universities will be able to field entries the second time around. So will hundreds of independent teams.
• People will develop solutions to some of the problems seen in this race. For example, one entrant got tangled up in barb wire (that could have happened to a human driver too -- barb wire is not so easy to see). But now the robotic vehicles will be redesigned so that they can detect and avoid barb wire.
• And so on.

So in 18 months, DARPA will run the race again, and lets say the winning contestant goes 14 miles. 18 months after that the winning contestant will go 28 miles. Then 56 miles, then 112 miles, and then finally a robot will complete the course. On that schedule, we have autonomous robots able to drive 150 miles on their own in just 7.5 years.

But the robots won't stop there -- they will keep getting better and better, just as computer chess machines did. In 15 years or so, robots will be driving trucks better than human drivers can. The robots will be safer and more reliable. They won't get lost. They will cause a lot less accidents. They will always follow traffic laws. They will never drive while drowsy or drunk. They will also cost a lot less than human drivers, and they will run 24x7. At that point, more than a million truck drivers will be out of work. The transition to robotic truck drivers probably happens in 2020 or so. See Robotic Nation for details.