@ the same mountainbike - I don’t know about that - the history of information devices has been a chain of things that “computers / robots / whatever we’re calling it” couldn’t do that humans could do, that suddenly - computers could do and do better than humans.
I mean, computers can read well enough that we have that crazy hard for humans to do CAPCHAs. They run the stock market faster and I assume better than humans did (for the financial companies anyway). They fly planes. They identify people in videos. They identify spoken commands. They parse e-mail for content. They can vacuum rooms.
Current systems like Eyesight can see cars and adjust the throttle to not hit them. I currently have this IN MY CAR. It’s not sci fi.
I just don’t see anything that makes driving more special than all the the above. We have lane following and adaptive cruise control TODAY. Google has self driving cars in prototype. Audi just had a car drive from San Francisco to Las Vegas.
Driving isn’t walking, and it isn’t balancing unless we’re talking about motorcycles.
And talking about more imputs - Humans have 2 eyes, pointed forward on their head. To look to the side means not looking forward. Cars already have 360 degree views in some Toyotas to help you park - adding additional cameras is easy, if potentially costly. Oh, and these systems could overlay radar, infrared, night vision AND the standard camera view. Robotic Cars have no reason to have a blind spot.
Humans can’t directly see in 2 directions at once, one radar and the other infrared heat. We have to look in one direction in visable light, perhaps at a screen that interprets the above, but we can’t act or interpret directly on it. We can’t feed it into our brain - but this is no problem for a robotic system.
I mean, computers can read well enough that we have that crazy hard for humans to do CAPCHAs.
You don’t know what CAPCHA is. The computer is NOT interpreting the capcha. It’s just displaying a predetermined string that’s distorted. The system cannot decipherthe capcha letters.
And then there’s reCAPCHA. Several companies are digitizing library books and the NY Times. The old type can be very difficult to decipher. That’s where reCAPCHA comes in. Ever see a capcha where you’re asked to type in two words? The first is capcha for security (Turning Test), the second is a pic from one of the books they’re trying to digitize which the system can’t decipher. So it’s asking humans to help with the second word/phrase.
They identify people in videos.
Watching too many CSI episodes. Identifying people from videos by systems is very unreliable. It’s one tool law enforcement uses, but it’s so inaccurate it’s not relied on very often.
I went to a seminar a few years ago for the telecom industry. One of the topics was Einstein and GPS. The accuracy of the clocks in the satellites relative to the ground clocks have to be extremely accurate. So Einsteins general relativity and special relativity theories come into play. The clocks in the satellites are moving much faster then the ground based clocks (special relativity), plus since the satellites are orbiting high above the gravity is 4 times weaker (general relativity),…so without calculating for Einsteins relativity theories the accuracy of GPS would be a couple miles instead of a few meters.
In 1997 there was a demo of self-driving car technology as well as “platooning” multiple cars close together to increase density. Carnegie Mellon University had a vision system that “drove” their car across the country 97% of the trip. It looked for lines, edges, gaurdrails or even the oil stain in the middle of the lane to keep the car in the lane.
The point is that this technology has been evolving for a long time and we are seeing the products hitting the road (well, maybe HITTTING is a poor choice of words…)
@jp10558: Your concept of a) robots “flying” planes and b) “driving” cars is, in a word, baloney.
Now, robots MAY “automate” the “time-and-motion” aspects of both activities…but the “executive functions” (i.e. Captaincy) remains FIRMLY in the hands of wetware, and will remain so.
Consider: in that plane “flown” by computers: who decides:
How much fuel to carry to account for an adequate reserve?
Whether or not to deviate from assigned route due to weather/turbulence/whatever?
If chest pain reported by passenger with cardiac history is sufficiently severe to warrant precautionary landing short of destination?
If the landing at the destination even should be attempted, in the face of Low-Level Windshear (LLWS) reports?
You cannot say a computer “flies” or “drives” ANYTHING until it substitutes for ALL of the functions previously preformed by the Human-in-Command. What you are describing, is an autopilot!
To say a computer can drive a car is just being simplistic. The CPU is but one part. The cog in this equation is them software and peripheral devices.
Can bugs be found and corrected which then in turns updates all other cars so they don’t make the same problem? Sure it can. But think about that for a second. It means that ALL similar vehicles will have the same problem.
I’m not sure AI to do this is there yet. There’s just so many variables. There will have to be YEARS of real world testing. And how do you do that easily without endangering everyone else on the road. Love to see how a system like that works in commuter traffic.
^My concerns (well aside from “captaincy,” which is more about the semantics of calling an autopilot “driverless”) is how well electronics work–or don’t work–in body-ground vehicles, subject to road salt, and with “typical maintenance.”
I’ve driven cars where the turn signals stop working when you step on the brake, where the coolant temp needle served as a metronome for the radio…and I had to scrap a car that kept intermittently going into “open loop” when I suspect (from experience after the fact) the problem was poor grounding.
Now, failure modes in such a rust-bucket is “interesting,” to say the least!
If something happens that causes the autopilot to “get lost” on the road, it no longer is 100% sure of its position on the road, it has but one choice. Maximum effort braking. Both the occupants of the car and any nearby drivers are certain to find this action unacceptable…
@MikeInNH I think you misunderstood me. The reason we no longer have the simple “Type in 3+4” style capchas is because computers can read and parse that and enter the answer. The reason we moved beyond an image of that is OCR is good enough to get it right enough that it doesn’t stop spamming. The capchas (like re-capcha) that work to keep out the spamming programs are hard enough that Humans have problems parsing them. I don’t know what we’ll do when we actually hit parity.
And I don’t think you’re arguing that computers won’t be able to drive cars eventually, you just think it’s more than 15 years out. Predictions are hard, especially of the future - but is Google and Audi and Volvo’s examples just dog and pony shows? I mean, did Audi lie about their car driving from San Fransisco to Las Vegas?
I just feel like I can hear the same arguments about automatics getting better fuel economy than manuals, or front wheel drive ever working right, or computers being able to generate music if I go back 40-70 years. In all cases it seemed unimaginable, yet it all happened.
@meanjoe75fan - you’re right, I do mean autopilot, though much of the captaincy issues I can see being solved after autopilot. I just think about Watson as an expert system for Doctors or the HFT systems on the stock market. You’d think you’d want humans making decisions for billions of dollars, but nope.
And I don't think you're arguing that computers won't be able to drive cars eventually, you just think it's more than 15 years out. Predictions are hard, especially of the future
I have a BS in CS and a MS in applied Mathematics plus 40+ years in the field. Computers and AI programming will eventually get to the point where they can drive a car or even do complex brain surgery autonomously…we’re not there yet. We have a long ways to go. I doubt I’ll see it in my lifetime. My grand kids might…but even that’s pushing it.
The current issue of Car and Driver has an editorial about test driving a new car with adaptive cruise control at 75 mph on an empty Montana interstate. He was completely relaxed behind the wheel, steering with one hand and his feet no where near the pedals when the car suddenly slammed the brakes on full.
He was so flummoxed by this that he got down to 20 mph before he could think to tap the brake to disengage the cruise control.
His conclusion was that increasing complexity always brings unforweeable problems.
At the obvious risk of comparing apples to oranges, I have to wonder about the mass use viability of computer driven cars given several instances in recent years of highest tech airliners crashing due to pilots not fully understanding how to effectively and safely manually fly the plane when a situation arose with auto pilot or auto landing malfunctioning?