Autonomous Audi TT Conquers Pike's Peak 187
fergus07 writes "After a year long research program, this week Audi revealed that its Autonomous TTS car had completed the 12.42-mile Pike's Peak mountain course in 27 minutes. An expert driver in the same car would take around 17 minutes — now we have a benchmark, the race is on, and it's almost inevitable that a computer will one day outdrive the best of our species, and it may be sooner than you think."
Bah! Stupid "the narrative" (Score:5, Insightful)
I suspect... (Score:5, Insightful)
Assuming that you don't totally cheap out on the fault tolerance or get horribly unlucky, the autonomous car should be able to complete the course every 27 minutes, with occasional pauses for refueling, and longer; but even more occasional pauses for hardware service on the car, virtually forever. That expert human driver, though, will do 17 minutes a number of times; but will be a downright danger to himself and others within 24 hours or so.
For many applications(municipal bus service and low-priority-low-cost mail delivery and commodity trucking/train deliver come to mind), it is virtually irrelevant what a top-notch human in fresh condition can do. What matters is either how many of those you can afford as spares, or what an exhausted, bored, hopped-up-on-stimulants-just-to-stay-awake human can do. Computers, on the other hand, may take longer than one would expect to catch up with best of breed humans in anything resembling natural conditions; but will be able to catch up with real world, performance-degraded humans considerably faster...
Re:The most important benchmark would still be... (Score:5, Insightful)
Something like a car, in a natural environment(not some empty closed test-track stuff), will suffer an accident from time to time. Human error, mechanical issues, sensor faults, algorithmic fuckups, whatever.
With a human driver(who basically all juries trivially recognize as an autonomous agent, since they think of all reasonably functional humans as such), the liability for accidents typically falls on one of the operators, unless a mechanical fault, braking issue, or something of that sort can be proven.
With an autonomous control computer, a jury will be much more likely to see the "driver" as an extension of the company who built the car, just like the brakes or the steering column, and assign liability accordingly.
Even if, lets say, computer-controlled cars delivered a 10-fold reduction in morbidity and mortality(which would save something like 35,000 Americans a year from death, plus an unknown but even larger number from serious injury, just to put things in perspective), that would likely be a net increase in liability for the vehicle manufacturers.
Until autonomous vehicles prove superior safety and insurers and/or legislators recognize the new state of affairs, it'll be strictly test tracks, tech demos, and internal use....
The best of us? (Score:3, Insightful)
I don't care about autonomous cars out-driving the best of us. I want to see common cars that can out-drive the morons on the freeways! Out-drive the mediocre and worst of us and I'd be happy.
Re:I suspect... (Score:5, Insightful)
but will be able to catch up with real world, performance-degraded humans considerably faster...
If you'd ever watched a semi truck driver cross the country non-stop running on speed (the driver, not the truck), you'd know it's possible to extend the number of hours a human can perform without significantly degraded performances. In my youth, I've always preferred hitching a ride with a truck driver than ride the bus, as I would invariably get there faster, and I never really felt unsafe.
As for the economics of autonomous vehicles, they'll become commonplace when
- a human behind the wheel is massively more expensive than the computer solution,
- people get over their fear of runaway machines,
- drivers unions are squashed
In short, it's not gonna happen anytime soon. Heck, even trains, the one kind of vehicle that could drive itself completely safely today, are still manned by "drivers" who spend their time pushing a button to tell the computer they're still alive, because passengers would be scared without drivers and unions prevent their removal from the trains.
Re:I suspect... (Score:3, Insightful)
The problem with using long-haul trucking as an example is that like so many jobs it consists of long stretches of boredom punctuated by moments of terror. You're either driving for long, boring periods of time, or you're having an "oh shit" moment due to equipment failure or driver error, usually the latter, and not necessarily yours. Why people like to cut off semis I'll never know. Even just hooking a travel trailer to the back of a 7 or 8 thousand pound pickup gives you lots of opportunities to have the same experience without getting paid. There is an answer to the problem that can be solved by machine, but it involves rails. Having robots drive trucks at this point would be dumb, but rebuilding the rail network where it's been neglected or dismantled and having robots drive trains makes perfect sense.
Re:Question though... (Score:3, Insightful)
The driver has to perform a mystical brain computation in order to integrate everything they're learning about the car through their five senses. The car can have as many senses as you have processing time and I/O to handle. The car can [theoretically] make more decisions per second as well. Ultimately the car is going to be faster... someday.
Re:Bah! Stupid "the narrative" (Score:3, Insightful)
Seriously, we could be looking at 50% unemployment in 20 years. It will be concentrated on the low end. How do we handle that as a society?
Try reading Oscar Wilde's "The Soul of Man Under Socialism". Over a hundred years ago he was imagining a world where machines were the new slaves and the majority of mankind was released from the drudgery of labour.