View Full Version : Man vs. Machine
Z-licious
03-10-2008, 08:30 PM
Me and James got into a discussion about computer drivers and human drivers today. It is my belief that a computer will one day will be able to take a car through a track faster than any human can take that car through the same track, as long as it's programmed well enough. My main argument is thus:
Once you break it down the basic level, racing is simply a matter of taking inputs from the car, processing the information, and outputting what needs to be done via the steering wheel, clutch, brakes, and throttle.
James believes there are some intangibles that would give the best human drivers an edge against any machine.
I would also like to open this up to say that anything that can be measured quantitatively, can be done better by a properly programmed computer. Anything that can be done competitively w/o the use of a judge (qualitative) will be the domain of the machine, if people go to the lengths to write the program.
Discuss!
Workdawg
03-10-2008, 08:38 PM
I agree with Eddy.
mazdamn02
03-10-2008, 10:48 PM
quantitative data wins. build computer eddy, gather data :)
dmention7
03-10-2008, 11:23 PM
Instead of speculating, why don't we program a computer to investigate whether a computer controlled car will ever beat a human. And then we can pit the computer's finding against leading automotive experts.
That will settle this once and for all.
mndsm
03-11-2008, 05:38 PM
Ultimately, the machine will only be as good as the man that builds it. Quantitative data aside, if the man programming the car is a gomer, the car is going to be a sled. Plus with machinery, there is absolutely NO way to program 100% of the variables into the car. At some point the machine is going to be asked to run outside of the programmed parameters. Man can compensate for that just by not sucking at driving. Machine will require more programming and even then, it's a crapshoot at best. I'd say at least in the forseeable future, man will be a better driver than machine.
Z-licious
03-11-2008, 05:59 PM
We are farther along then you think:
http://www.youtube.com/watch?v=AlYX02IrVbg
At the rate that technology is advancing I predict that we will see a robot driver that will beat any human within 20 years.
And about the parameters? A computer can accept an infinite range of parameters. But the amount that could happen on a track are limited, so you'd only have to program those, which would not be impossible.
mndsm
03-11-2008, 06:02 PM
It's the if/then that gets me. A computer certainly accept infinite parameters, but can man program them? Because one little miscalculation with programming and the computer is screwed. Look at Windows.
Z-licious
03-11-2008, 06:09 PM
Windows is a robust operating system designed to handle a myriad of applications and hardware. A robot driver would be one application programmed to run on one piece of hardware. When you specialize like this, you don't see the kind of crashes you run into with a full blown OS.
dmention7
03-11-2008, 06:33 PM
The thing you guys aren't considering is the approach that will likely end up being applied to most problems that involve extremely complex inputs and a correspondingly complex response to those inputs.
Genetic algorithms (or some variant on that theme) allow a program to "evolve" to optimally solve a task. Aside from writing the code that frames the problem, and allows the program to evaluate its progress, these programs write themselves with minimal human intervention.
They don't necessarily require any prior knowledge of how to solve the problem to arrive at algorithms that are more efficient than a solution that was engineered by a human. It's a lot like how the way nature solves a biological problem is often quite different than how an engineer would solve the same problem.
Anyways, the basic idea is that developing a program to drive a car better than a human would not necessarily require the programmer to have the knowledge to anticipate and program for every possible combination of variables. A learning program exposed to a real-world dataset would be able to make the necessary connections itself much more efficiently.
Z-licious
03-11-2008, 06:47 PM
Yes, very good point. A learning program would take much of the weak human element out of the equation, and again, would do it much faster than a human.
So, not only can a machine drive faster than a human, it can learn how to drive faster than a human as well!
Big Nate
03-11-2008, 06:50 PM
Well I agree with Jay. (he is good at talkfighting)
Fuck robots, no really, why waste time on driving robots if we can make life like bots to have sex with.
Oh shut up, you know you thought about it. ;) LOL
BTW, that video was awesome!
Powered by vBulletin® Version 4.1.5 Copyright © 2012 vBulletin Solutions, Inc. All rights reserved.