The DARPA Race

They ran the first DARPA race yesterday. Not surprisingly, given the freakophile nature of the press, most media treated this as a real-life version of Wacky Races.

For instance, here (free reg required), here, and here.

They don’t get it, just like most people don’t. And this is why they fail at many things in life, because they don’t understand the nature of failures, and how to deal with them.

With the right approach, failure is constructive, even necessary for success. With the wrong approach, failure is just . . . failure.

Getting a vehicle to drive 150 miles through nasty terrain is hard, and this basically was the first trial run. Computers work well where the rules are simple and “can’t handle it, give me something better to work on” is an acceptable answer. “Can’t handle it” is not an acceptable answer for an unmanned vehicle (at least not very often). Those first vehicles were basically asked to walk and chew gum at the same time for the first time, and that’s really, really hard for a computer to do.

Come to think of it, it’s not so easy for people, either. When you were infantile, did your parents dump you off in the Mohave Desert, give you a car, go 150 miles away, and say, “Come to Papa?”

Did they do it when you were five? Or eight? Or twelve? At least in America, the law doesn’t allow you to drive on roads until you’re at least sixteen.

Even if you’re a good, even great driver, if you were given this task, you’d probably fail the first time. You’d probably not make some of the mistakes these robots made, but you’d get lost or break an axle, or get stuck.

However, if you did it enough, in all likelihood, you’d eventually get the hang of it, simply because you’ve learned by your earlier mistakes.

But that’s not necessarily the best way to learn. If you got yourself a book of how to drive an off-terrain vehicle, and what to do in common situations, I bet you’d figure it out a lot quicker, and avoid many mistakes.

Why would it be any different for a robot? Especially one that has rather less intelligence than that of a bumblebee?

Lessons To Be Learned

From the results of this first race, there are two types of lessons to be learned.

First, those teams learned things that could not have been reasonably anticipated in advance, and will build them into their future models.

Second, at least some of the teams learned things that could have been reasonably anticipated in advance by at least somebody, if not the team members themselves. For instance, somebody used to driving an off-road vehicle probably could have given helpful hints on what to do or, more importantly, what not to do to even the most serious of teams

Different people will learn different lessons from this.

Some will give up, for different reasons. Some of them will give up just because they didn’t succeed the first time. Some others will give up because the problem is bigger than their time, effort and budget will allow; the only question for them was whether they could have figured that out before they tried the first time.

Some won’t learn from this, or acknowledge that there was a mistake. They were right; it was reality that was wrong, or unjust or whatever. They’ll just repeat the same mistakes the next time. Some will be minor, some will be fundamental.

Some will learn from their own mistakes, but not the mistakes of others, and repeat the others’ mistakes the next time.

Some will learn from all the real-life mistakes, but won’t do the homework to find out about the mistakes no one has run into quite yet. They’ll take care of that omission when they fail the next time.

The smartest will learn all they can from the books and from others. That doesn’t mean they’ll succeed; it will mean they’ll have a good failure the next time around.

A “good” failure? Why, yes. There are good failures and bad failures. A good failure happens when you fail because you’re breaking new ground, and you learn something you couldn’t have learned any other way. A bad failure is when you stumble on old ground, or at least one that has a map.

Those of us who overclock can learn some lessons from this. Our task is far simpler than those competing for the DARPA awards, but even the smartest of us have fallen into one of the mental quicksand traps described above.

It would have been best to do it right the first time, but if that didn’t happen, it’s better to do it right the second, or fifth, or tenth time.

What is the “right” way? Learn what it takes to do it right before you do it, be ready to do whatever it takes before you do it, or accept the results if you’re not, and realize that even after you’ve done all this, you may still fail.

Unless we set the pole vault bar very low, success isn’t never guaranteed. All we can do with our efforts is improve the odds for success in our favor, and if we fail, have “good” failures.

If you’re not willing to do that for yourself, if your “work” substitutes hopes and/or prayer rather than work; you’re kind of like the entrant who spent less than $20,000 but had two Buddhists monks bless the entry.

It never got out of the starting gate.


Be the first to comment

Leave a Reply