Mobileye's Self Driving Car is unbelievably good

[Bladerskb]

Overwhelmingly impressive. Shows you how ahead they are of the competition.

Article

How Intel and Mobileye are taking a simpler path to autonomy - Roadshow

(Extra Clips) Handling difficult Lane Changes and Merges

Start at 33:00 mins

 

[malcolm]

Not really a "simpler path to autonomy":

But at a press event in Mobileye’s hometown last week, a car outfitted with television cameras from Israel’s Channel 10, went straight through a red light about a quarter of mile from the company’s garage after an otherwise uneventful ride.
Nobody was hurt, and Channel 10’s video seems to show a Mobileye safety driver monitoring the vehicle, but allowing the car to proceed without trying to stop it. Mobileye’s Chief Executive Officer Amnon Shashua said wireless transmitters on cameras used by the television crew created electromagnetic interference, which disrupted signals from a transponder on the traffic light. Although the car’s camera correctly identified the light as red, the car ignored that information and drove based on signals from the transponder, a mistake that has since been corrected, he said in an interview

Bloomberg

Claiming that the combination of cameras + lidars + radio signals is bound to be safer doesn't really hold up in this example.

Why didn't the input from the cameras override the lack of radio signal? Isn't that the whole point of multiple redundancy?

Has correcting for this error introduced a bias which will cause problems in another situation?

TLDR: Plenty of sensors, not enough brains.

 

[Superendo]

Imagine what would have happened if Tesla and Mobileye had been together working to the FSD goal.

Although very impressive demo it is not fsd No pedestrians bicycles and wierd road infrastructure etc etc

 

[ChrML]

Overwhelmingly impressive. Shows you how ahead they are of the competition.

Article

How Intel and Mobileye are taking a simpler path to autonomy - Roadshow

(Extra Clips) Handling difficult Lane Changes and Merges

Start at 33:00 mins

Good presentation. It explains the "common sense" of how self driving vehicles will work and the best way of approaching the problem.

There is nothing new in the video that Tesla or other developers of self driving cars don't know. But somehow if Tesla was this mature in their development, then Elon would show it to the world... That's why I believe MobileEye is much further ahead.

 

[croman]

Bring this tech into real cars and I'll be more impressed. A lot of those cameras are poorly integrated and honestly the rest seemed to mimic Tesla's design. Otherwise Tesla's propaganda has soured me enough on other "demonstrations" to ignore the hype until its demonstrable to me in person.

 

[Cosmacelf]

I agree, very impressive. I’ve always thought Mobileye was staying ahead of Tesla. Elon totally underestimated what it takes to build a self driving car. Tesla is now about even with what Mobileye had several years ago. However, I don’t know if Elon had a choice since Mobileye was freaking out that Tesla was using their technology the way they did. Personally. I’d give it 50-50 chance Tesla ends up using Intel self driving technology in a future car.

 

[Tam]

...Overwhelmingly impressive..

Construction zones, traffic cones, laneless roads, dirt roads... are very difficult scenarios that Nvidia showed on its demo 2 years ago with 1 single frontal camera and no other sensors:

 

[Wooloomooloo]

Not really a "simpler path to autonomy":



Bloomberg

Claiming that the combination of cameras + lidars + radio signals is bound to be safer doesn't really hold up in this example.

Why didn't the input from the cameras override the lack of radio signal? Isn't that the whole point of multiple redundancy?

Has correcting for this error introduced a bias which will cause problems in another situation?

TLDR: Plenty of sensors, not enough brains.

Someone has to say this - that's fanboy talk. "Simpler Path to Autonomy" is obviously just marketing, but there's no doubt that their autonomy system is far ahead of anyone else (including Tesla).

You know you CAN love Tesla and appreciate a competitor product as well.

FSD is still half a decade away, minimum.

 

[TaoJones]

Someone has to say this - that's fanboy talk. "Simpler Path to Autonomy" is obviously just marketing, but there's no doubt that their autonomy system is far ahead of anyone else (including Tesla).

You know you CAN love Tesla and appreciate a competitor product as well.

FSD is still half a decade away, minimum.

Well said. Especially since Mobileye is part of an increasingly crowded Driver Assist field.

Many many people will be perfectly content to buy a new version of their ICE if it has “almost as good” DA features. Already that marketing/advertorial effort has materialized (see Nissan’s ProPilot).

I don’t know that people in general (and fanbois) have fully grasped that today’s AP1 and even today’s lesser AP2 will be features for the most part available from every brand in 1-3 years.

I hope Tesla continues to win subsequent stages of this vehicular Tour de France. But it is by no means certain. Note the recent claim that the new Tesla SoC is 10x faster than NVidia’s SoC. You have to dig just a little deeper to reveal that the sentence should continue, “ currently in Tesla vehicles.” *Today’s* NVidia product *is also 10x faster.”

Sometimes the whole truth hurts, but it is what it is.

You know the nature of AI... muddle, muddle, muddle, and then BAM. Or voila, if you prefer. Thing is, everybody’s doing it. I expect in the future that we will see leapfrogging much as we saw with cell phones. One manufacturer, then the other will have the better product, each obsolete before it hits the market.

Even there, Tesla has an inherent edge twice: First, they’re not a car company at heart - an obvious positive. And second, they aren’t tied to annual “new models”.

I like their chances but I’m no longer willing to bet the farm on it until they actually do what they say they’re going to do when they say they’re going to do it. Without ambiguity and without the inevitable class action lawsuits.

 

[traxila]

Someone has to say this - that's fanboy talk. "Simpler Path to Autonomy" is obviously just marketing, but there's no doubt that their autonomy system is far ahead of anyone else (including Tesla).

You know you CAN love Tesla and appreciate a competitor product as well.

FSD is still half a decade away, minimum.

Let’s see what happens with the new chip.

 

[strangecosmos]

What am I missing? The Roadshow video just looks like a demo of autonomous driving on a highway. Just doing highway driving is a lot easier than what we’ve seen in demos from Waymo, GM/Cruise, Tesla, Zoox, etc.

The video with Amnon Shashua is cool because it shows an autonomous car assertively merging in heavy traffic on a highway in Israel, where drivers are more aggressive than in the U.S. or Canada. That’s something I haven’t seen before.

 

[mongo]

Even there, Tesla has an inherent edge twice: First, they’re not a car company at heart - an obvious positive. And second, they aren’t tied to annual “new models”.

Third, they already have a fleet of cars with all the sensors on then for data collection and evaluation (not making any claims on the level thereof). So many (potential) test drivers, so many (potential al) situations, so much (potential) data.

 

[Bladerskb]

What am I missing? The Roadshow video just looks like a demo of autonomous driving on a highway. Just doing highway driving is a lot easier than what we’ve seen in demos from Waymo, GM/Cruise, Tesla, Zoox, etc.

The video with Amnon Shashua is cool because it shows an autonomous car assertively merging in heavy traffic on a highway in Israel, where drivers are more aggressive than in the U.S. or Canada. That’s something I haven’t seen before.

Let's be honest, you cant put tesla in the same sentence as these guys. Tesla demo is based off a 1 mile per disengagement system. And the demo was extremely easy, empty road basically. What ME demonstrated is FAR more advanced.

 

[strangecosmos]

What ME demonstrated is FAR more advanced.

Why is Mobileye’s demo more impressive than demos from Waymo, Cruise, or Zoox? It’s just highway driving. Cruise’s demo is in downtown San Francisco.

 

[VT_EE]

Let's be honest, you cant put tesla in the same sentence as these guys. Tesla demo is based off a 1 mile per disengagement system. And the demo was extremely easy, empty road basically. What ME demonstrated is FAR more advanced.

Haha. Reading your posts is like watching Fox News. I know exactly the angle you’ll take regardless of the topic. Always anti-Tesla. You and Einhorn are great for each other.

 

[strangecosmos]

Here’s a demo video from Drive.ai that shows a car operating autonomously in a more complex environment and heavier traffic than the Roadshow video:

 

[SR22pilot]

All of these controlled demos are nice. However, why is Mobileye's significantly better than this:

I want to see side roads, stop signs, traffic lights etc. Now the nVidia demo with road work is a different story.

 

[Bladerskb]

Why is Mobileye’s demo more impressive than demos from Waymo, Cruise, or Zoox? It’s just highway driving. Cruise’s demo is in downtown San Francisco.

I don't understand how anyone could read my post and come out with the same conclusion. I think i made it pretty obvious i was only talking about the Tesla demo.

 

[Bladerskb]

All of these controlled demos are nice. However, why is Mobileye's significantly better than this:

I want to see side roads, stop signs, traffic lights etc. Now the nVidia demo with road work is a different story.

You mean a footage that has about 8 turns on public road (7 from stop sign), (1 from traffic light). No lane changes. These are the things mobileye been doing since 2013. These are easy. The footage is also littered with mistakes.

The rest of the footage from :46 secs to 2:30s is footage of driving on the road with absolutely no car in sight. zero. none. And that's supposed to be impressive? That's like 5 miles of driving on completely empty road with no road users.. Mobileye is doing something that haven't been done before. I have only seen cruise do on a more cautious level.

If all you want to see is side road, stop signs, traffic lights, then go watch videos from Delphi (aptiv), and Nissan. They both use mobileye for vision and mapping. Personally, i don't really care for demos and only care about production launches.

 

[Bladerskb]

Haha. Reading your posts is like watching Fox News. I know exactly the angle you’ll take regardless of the topic. Always anti-Tesla. You and Einhorn are great for each other.

Watch this below and reply to me a video of a car doing this.

Intel and Mobileye Autonomous Test Car Drives in Jerusalem from Intel PR on Vimeo

The only other people doing stuff that is equally impressive although not exactly the same thing is cruise and waymo.

@Trent Eady