Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

"Look Who's Driving" (Nova episode)

This site may earn commission on affiliate links.
I hope this is not a duplicate post. I didn't see it anywhere.
Last night there was an interesting episode about autonomous driving and of course Tesla was mentioned in there. It is supposed to air again on Friday and on Sunday and I think it is on most PBS stations, DirecTV, etc. Worth watching, but I don't think you would learn much more than you already know. It's informative for new owners whom might not be as familiar with the state of the art.

So as the tile implies, on Nova, this episode is called "Look Who's Driving"
 
  • Informative
Reactions: mswlogo
If you don't mind streaming, you can watch it any time on its website:

Look Who's Driving

Also, it's been mentioned in:

Autonomous Car Progress

I find it interesting that they give the last word to a PhD who says the tech isn't ready, yet Waymo is now actively using non-safety driver cars. It's so hard to tell how this will play out. Good video though. Mobile Eye still has the best 'aggressive' lane change stuff in the business ;)
 
I find it interesting that they give the last word to a PhD who says the tech isn't ready, yet Waymo is now actively using non-safety driver cars. It's so hard to tell how this will play out. Good video though. Mobile Eye still has the best 'aggressive' lane change stuff in the business ;)


Machine works best when the rules and environments are all finite and static.

That means it can play chess fine because there are only finite moves.

It can operate the elevator fine because of the same principle that all rules and environments are codified.

Waymo uses that same principle and says what if the environment is nonchanging for its machine? That's why it is geofenced. It would prepare the roadways meticulously: where are the traffic lights, where are the traffic signs, where is every tree, every light post...

It makes sure everything in the geofenced zone is the same so its machine can work better.

There's nothing wrong with preparation.

However, it becomes impractical because if I want to use Waymo's technology, I have to move from California to the geofenced area of Arizona.

Thus, the technology is ready if you play by its rule: the chess rule, the elevator rules: all moves and obstacles are finite and predictable.

However, the technology is not ready if you play by human rule: In Los Angeles and Bakersfield, the traffic is temporarily redirected to drive on the opposite side of the freeway (physically wrong roadside for hours or for months) and with all the sophisticated of GPS, Waymo's HD pre-mapping, it just can't handle it!
 
I find it interesting that they give the last word to a PhD who says the tech isn't ready, yet Waymo is now actively using non-safety driver cars. It's so hard to tell how this will play out. Good video though. Mobile Eye still has the best 'aggressive' lane change stuff in the business ;)
yes, I got a laugh on that one. He said level 4 and 5 isn't going to happen for a very long time..... a vague reply. Just how long is "very long".
 
yes, I got a laugh on that one. He said level 4 and 5 isn't going to happen for a very long time..... a vague reply. Just how long is "very long".

I've seen "experts" say it will be 30-40 years before we get to L5 autonomy. I can only assume that they mean L5 that is "perfect", ie works everywhere, non geofenced and has perfect safety and reliability. Yeah, it might be decades before we get that level of safety in autonomous driving. Granted it is geofenced, but we are seeing legit L4 robotaxis pop up more and more. So L4 on a small scale is starting to happen even now.
 
  • Like
Reactions: Silicon Desert
The video mentions that despite the tens of thousands of fatal accidents each year, statistically, humans are actually incredibly safe drivers. There is only 1 crash death per 100 million miles. The narrator makes the case that it is an incredibly high bar for autonomous cars to meet and expresses skepticism that autonomous cars will be able to do it. Certainly, if you want autonomous cars to be true L5 where they can drive anywhere and everywhere, billions of miles per year like humans, and be safer than humans, then that standard is correct. And yes, it may be awhile before we achieve that standard. But I think it ignores that there is a way to make autonomous cars safe enough and useful and practical sooner.

If you restrict autonomous cars to a smaller, geofenced area, and thereby make it a more predictable and controlled environment, you can dramatically reduce the risks of accidents. By doing this, you can increase the safety and reliability of the autonomous cars in the area that they operate in, thus allowing you to deploy your robotaxis sooner. In fact, this is what Waymo and others are doing now: using HD maps, an overabundance of sensor redundancy and relatively small geofenced areas, in order to maximize the reliability of their robotaxis where they can be deployed in their geofenced areas even before they reach the 1 crash death per 100 million miles standard.
 
...yet Waymo is now actively using non-safety driver cars...

Almost 2 years ago in January 2018, Waymo announced that it's confident that its new Waymo One program to the public would not have a safety driver. But it abandoned the idea and still has backup drivers all this time.

Now, on Sunday, Waymo said it has begun driverless rides with no backup drivers to a few hundred users but they also signed non-disclosure agreement.

With a gag order, it's hard to know how well the program is.