Tuesday, August 7, 2012

Humans and their robots

Having Curiosity land on Mars has sparked some thoughts. I am still amazed that the landing worked. When I saw the first concept film of what the landing would look like, I thought they were out of their minds. Everything would have to go perfectly, and there are nearly an infinity of things that could go wrong.

Some people say the space program was a waste of money. That's the subject of a good rant right there, but I'll content myself by saying "medical diagnostics", and remembering what my Grandad taught me about arguing with idiots.

Since the beginning of recorded history humans have experimented with making things. Then when it broke they would figure out why it failed, and try again. And again and again and again as long as the religious fruit loops or threatened aristocracies didn't stop them. That process worked brilliantly with aircraft, which is why it's one of the safest ways to travel. Far, far safer than by automobile. (I'll get there.) Figuring out why an airplane crashed usually involved going to pick up the pieces, getting data from the flight recorder, and statements from witnesses, some of whom were going to die very soon and yet were still trying to make the plane fly while talking about what was happening.

Then a few decades ago we started playing with rockets. Failure meant a loud noise and shower of metal confetti. The only way to figure out what was happening was to put telemetry on the rocket and have it send signals back to earth. We take it for granted now, but it's a hard thing to do.

Then we wanted to put people at the top of those rockets and send them to the moon. Failure was no longer just an expensive fireworks show. It would kill valuable, highly trained, extremely skilled people that were very difficult to replace. It was imperative to figure out how to build a rocket so that everything would work properly, first time, every time. Every component, every rivet, nut and bolt, screw, wire, switch, circuit, sensor, everything had to work. It took quality process and management to an entirely new level. This sort of work is one of the reasons are cars are so much safer now. (I'm getting there.)

Then we wanted to put instruments on Mars. A difficult target. Let's think about Spirit and Opportunity. They were designed with the intent to last 90 sols. Spirit lasted 2210 sols. To save you from doing the math, that's nearly 25 times the design life. Opportunity is still going strong, more than 3100 sols later. That's more than 34 times design life. If only our cars could do so well. My buddy Leauxra sent me a link to a cartoon about Spirit. We can't help but humanize our creations, and I'm sad that Spirit is no longer functioning, but proud that it lasted so much longer than we thought it would. Eventually humans will get to Mars, and little Spirit can be put in a museum.

Curiosity has some big shoes to fill, but the mission is starting out well. With human and earth computer help it flew to a tiny square at the top of the atmosphere, then on it's own it flew down to a safe landing within 2 Km of the aiming point, after more than 350 million Km of travel. That's about 35,000 times more accurate than a human aiming for a parking spot downtown, and I have daily evidence that some humans can't manage it.

Which brings me to a news item I saw recently. Google just passed 300,000 Km of testing a driverless car without a single collision. The only time a car with the driverless system was in a collision, a human was driving it. This is better than the average American driver.

Now, I have to digress for a brief moment. I'm death on people saying "car accident" in my hearing. I won't put up with it. The word accident implies that there was nothing that could be done, and that is BULLSHIT! Almost every "accident" is attributable to driver error. Some driver screwed up, either by not paying attention to the task at hand, or by being under the influence of drugs, or driving beyond the limits within which they could control the car which under most circumstances is another failure, since the car is probably still controllable by a real driver. These collisions cost our economy billions and cause untold heartache from untimely deaths and injuries.  

The sooner we stop allowing humans to drive, the better off we'll be. Back to Google. Driverless cars have been a staple of science fiction since the early days. Now Google has done it although there is still a long way to go before its commonplace. They have to teach the system how to deal with snow, and construction detours, and no doubt how to deal with the remaining dangerous cars on the roads, the ones with a human at the wheel. A human that believes that he (almost certainly a he) can drive better than any computer (he can't) and that nobody can take away his God given right to drive (we should, preferably now).

They will base their emotionally over wrought argument on obscure and unlikely circumstances where a prepared human might do better than a computer. Maybe. I suspect that the computer won't be taking risks that lead to those circumstances in the first place, and is far more likely to take corrective action much sooner than a human could. I remember reading about someone ranting about daytime headlights.  They were incensed by the thought that they couldn't control the lights, and thought that they should be able to control every aspect of the car's performance. Naturally, they thought automatic transmissions were the work of the devil. I wondered if they wanted to manually control the subtle and complex interactions between fuel and air supply that happen many times a second to make the engine produce the required amount of power.

There is no shortage of evidence that most drivers are not up to the task of safely driving an automobile, even when they are in the so called prime of life. Don't get me started on geezers. Everybody says that more frequent testing should start at an age about 10 years older than they are. I say, everybody should be tested more frequently, and be tested much more rigorously, preferably using car simulators to save the carnage as they fail. My license renews every 5 years, (next in 2015), and I'd be perfectly happy with every 3 years, with a test. My question is, if you know what you're doing, why fear a test?

I'm pretty sure that a computer driven car will behave more predictably than a human driven car. I'd like to find out if all the processing happens on board. I suspect so given that some of the testing territory is pretty remote. Can't wait for it to happen.

Another example of a really simple robot. The spit on my BBQ. It worked for lamb roast tonight.


Very yummy! No formal workout today, just a good stretch session. I can now touch my toes again and relax into it! Yay! One side is still tighter than the other, but my back is almost back to normal. Can you say weekend bike ride? Core included front, side, front, side, front, side, front, side, front plank.




2 comments:

  1. I love this post! I am a very good drive, and I HATE driving. If we can't have teleporters, I want to lounge in a seat and READ while I commute, not DRIVE. And I hate how NASA keeps getting gutted. Some perspective: Cost of Curiosity Program: MSL project is about US$2.5 billion Cost of Stealth Bomber program: US$44.75 billion (through 2004), US$2.1 billion per unit Guess which one I would rather we built?

    ReplyDelete
  2. Thanks for loving this post! Did you know you can subscribe to a twitter feed for Curiosity? I did. That's almost enough for 17 siblings for Curiosity. I know which I'd rather have too. When you think that Mars has almost the same surface area as earth's land mass, and think of all the odd corners that we've barely explored, and how we keep finding stuff, there is lots on Mars for the rovers to look at.

    ReplyDelete

Looking forward to reading your comment!