Mustang v. Plaid

[andrewtac]

Correct and there will be negative outcomes regardless of how good it is.

The unknowns can be reacted to not predicted. Just like we drive today, we take our path and then some jerk interrupts our plan and we react. Even some of the really basic stuff on cars now do this, the assisted stop for example. Far from perfect, but it is an example of the car reacting to external stimulus.

That technology can cover all known situations. The biggest problem, though, are the unknown ones.

Sponsored

 

[wynand32]

That technology can cover all known situations. The biggest problem, though, are the unknown, unpredictable ones.

And even in known situations, it's not smart enough yet and won't be for a long time. I'll refer again to our debate about assisted braking.

 

[andrewtac]

There are radars now that can sort hundreds of thousands of threats simultaneously from very long distances, they can make images as good as pictures through weather from a long ways away.

 

[andrewtac]

And even in known situations, it's not smart enough yet and won't be for a long time. I'll refer again to our debate about assisted braking.

Assisted breaking is an example of reacting to an unknown. It is an extremely simple sensor and process, extremely far from what is used in other platforms.

 

[Vlad Soare]

Assisted breaking is an example of reacting to an unknown. It is an extremely simple sensor and process, extremely far from what is used in other platforms.

And yet it can't even do its simple job properly. Let alone work continously in all imaginable situations.

 

[wynand32]

Correct and there will be negative outcomes regardless of how good it is.

The unknowns can be reacted to not predicted. Just like we drive today, we take our path and then some jerk interrupts our plan and we react. Even some of the really basic stuff on cars now do this, the assisted stop for example. Far from perfect, but it is an example of the car reacting to external stimulus.

Yes, technology can respond to external stimuli. Even simple systems can respond to external stimuli, like a thermostat that adjusts a heating/AC unit based on temperature. But it's not the 90% or whatever times that more complicated systems get things right (and that's probably being generous) that matter; it's the 10% or whatever times that such systems get things wrong that matter. When every car is making poor decisions, they get compounded.

AI just isn't that smart.

 

[Mike Pfeifer]

The only way self driving will ever be successful is if its like a Disney ride where all the vehicles are controlled using the same system within all the rules. You cannot have even one human driver mixed in or it all falls down. You have to take all the unpredictability out of it.

 

[wynand32]

Assisted breaking is an example of reacting to an unknown. It is an extremely simple sensor and process, extremely far from what is used in other platforms.

And yet it can't even do its simple job properly. Let alone work continously in all imaginable situations.

Exactly. If we can't rely on something so "simple" to work consistently enough to be generally applied, then we certainly can't expect even more complex systems to work consistently. All the problems that exist with the simpler implementations exist in the more complex implementations, only magnified and multiplied.

 

[Vlad Soare]

There's a reason why pilotless airliners aren't allowed, despite having the technology to fly themselves. And that's technology that actually works! Even that isn't deemed safe enough, and for good reason.
Anyway, I agree with Clarkson. Until someone sits in a self-driving car all the way down the Death Road, the point is moot.

 

[wynand32]

The only way self driving will ever be successful is if its like a Disney ride where all the vehicles are controlled using the same system within all the rules. You cannot have even one human driver mixed in or it all falls down. You have to take all the unpredictability out of it.

Yes, this is precisely right. One system controls every vehicle on the road so that everything is not only predicted but tightly controlled. Maybe someday we'll create AI with general intelligence and it can perform like the human brain in such scenarios, but that's an exceptionally long time away if it's even possible.

 

[MAGS1]

Yes, this is precisely right. One system controls every vehicle on the road so that everything is not only predicted but tightly controlled. Maybe someday we'll create AI with general intelligence and it can perform like the human brain in such scenarios, but that's an exceptionally long time away if it's even possible.

I don’t know about anyone else, but I doubt anyone can make AI replicate the way my brain works (or doesn’t work) . Just injecting a little Monday humor, continue on, I find this discussion quite fascinating

 

[Mike Pfeifer]

I don’t know about anyone else, but I doubt anyone can make AI replicate the way my brain works (or doesn’t work) . Just injecting a little Monday humor, continue on, I find this discussion quite fascinating

That makes two of us!

 

[wynand32]

There are radars now that can sort hundreds of thousands of threats simultaneously from very long distances, they can make images as good as pictures through weather from a long ways away.

The sensors aren't the problem. It's the intelligence necessary to make use of them.

There's a growing belief in AI circles that it's impossible for us to use contemporary methods to create general intelligence, because it can't be mathematically modeled in such a way that algorithms can be written to accomplish it. And it's something awfully close to general intelligence that's needed for something like driving a car on open roads.

I wish I had more time to delve into it here, not that I wouldn't just bore the heck out of everyone. But it's a pretty complex topic that very few people understand (including me), and again, those who do understand it -- if they're honest -- are pulling back on sensationalist claims.

Note that Elon Musk is a bit of a freak in this regard. He has some wacky ideas, including the notion that we're probably all living in a computer simulation. You can't believe something like that is even possible without accepting some pretty unrealistic (i.e., irrational) premises.

 

[RagmopInKona]

Just one more reason to dread a discovery of the fountain of youth.

I'll take that risk..lol

 

[wynand32]

I don’t know about anyone else, but I doubt anyone can make AI replicate the way my brain works (or doesn’t work) . Just injecting a little Monday humor, continue on, I find this discussion quite fascinating

You know, you say that in a joking way, but you're not far off. The fact is, nobody has the slightest clue how our minds work, so replicating them with a computer is impossible from that fact alone. And the more we do understand, the less it seems like a computer can do what our brains do.

Simply put, unlike what has been assumed for a few decades, our brains aren't just computers running software. There's something else going on that can't be mathematically modeled in a way that computer algorithms can duplicate.

Sponsored