When do you plan on making this point?Erm...lemme think. Chicken Little led a futile existence ?
When do you plan on making this point?Erm...lemme think. Chicken Little led a futile existence ?
LOL. Woosh!When do you plan on making this point?
But, it would be great for when I get old and they take my drivers license away....
<SNIP>....
Would you like some help?LOL. Woosh!
I work in a beverage can plant, and the vision system to detect defects, in cans, runs at over 2000 cans a minute and can reject a can for a pinprick sized dent, or off color, without missing a beat.I know - I've built machine vision inspection systems. It's not a trivial problem, no question. But the technology is advancing quickly.
That was just an example. Now, consider machine vs human for a moment. A human can look in one direction at any one time, and can scan a reasonable amount over a short period of time. But there are blind spots, and always the chance of somebody creeping up on you in a blind spot while nobody is looking. And that's in good conditions, when you're alert, etc.
In an automated car, you've got 360 degrees of coverage, fully alert in all directions all the time. Sorry, but no human can compete with that.
The technology is already pretty good, and only going to get better.
bs
I bet there's a lot it doesn't do "yet". Any links?We have just delivered a tractor trailer of beer across the country driver-less to the loading dock.(does not back into the dock yet)
OK cool - that kinda puts it into some context.I was a little mistaken with the distance,not quite cross country.
https://www.bloomberg.com/news/arti...th-budweiser-makes-first-delivery-in-colorado
http://www.cnbc.com/2016/10/25/driverless-beer-run-bud-makes-shipment-with-self-driving-truck.html
Computer processing is all about doing a lot of simple things very quickly. You can increase the number of things to do and you can increase the speed but you're still stuck in this groove of sequentially processing a bunch of simple things quickly. Visual data is fed into this groove in a raster sequence and contextual / situational awareness is gradually built up to some threshold where a decision is made regarding the control of the vehicle - ignoring hard interrupts which force "fail safe" behavior.Not at first but when they are proven to be significantly safer than humans yes. They will become safer as more of them are on the road. I think that when I am old, the majority of cars will be self-driving.
Today machine vision is way more advanced than machine perception. Simple creatures that evolved millions of years ago have better instantaneous seven-dimensional (x, y, z - height, t - time / motion, RGB - color) perception than today's machines. So - if we're waiting until the algorithms produce a level of machine perception equal to that of simple life forms - let alone that of humans - IMHO we have a very long wait ahead of us.
That's like saying that life is mostly the five senses - and you really don't need a brain as long as your nose works, your ears work, your eyes work, your taste buds work, your nerves work. Very simple organisms bring more to the table than that. And they've benefited from millions of generations of genetic algorithms.It doesn't require the entire perceptive capability of a human to drive; it mostly just requires that the sensors are operating.
That's like saying that life is mostly the five senses - and you really don't need a brain as long as your nose works, your ears work, your eyes work, your taste buds work, your nerves work. Very simple organisms bring more to the table than that. And they've benefited from millions of generations of genetic algorithms.
Bees do this little dance that tells the other bees where the flowers are. But algorithmically that's probably nowhere near as sophisticated as visual perception. For a smart car sensor to have true depth perception - not radar ping ping ping - it would need two separate cameras and the brains required to process parallax virtually instantly across the field of view. Assuming it could accomplish this then it would need further brains to interpret this processed depth perception.Fairly sophisticated behavior can be driven by minimal processing. For instance, the foraging behavior of social ants can be generated using an algorithm that only has 3 or 4 decision rules.
Would you mind making such an argument? I realize it's easy to be impressed by modern technology cuz of smart phones and computer generated special effects in movies. But these were easy problems for which the answers were already known - all that was needed was the for the hardware to evolve to the point where the implementation of these answers became practical. The answers for the problems of self-driving cars are not dependent on the evolution of hardware - they are dependent on the evolution of algorithms.I think you are making this more complicated than it is. Algorhythm shmalgorhythm. I could argue your "evolution vs algorhythms" argument is irrelevant to the issue at hand (safe self-driving cars).