As a Tesla shareholder, I am surprised to see a competitor that somehow inspires me more confidence in what they are doing. Precise information rather than overwhelming non- backed up promises. My prediction would be that while MobileEye probably will have an edge in full autonomous driving, the vertical integration will Tesla allow to compete successful. That is, if both companies self driving systems will be authorized. If only one makes it onto the streets, that company will dramatically win market share. In the case of MobilEye this means a number of OEMs plus Nio plus their own Robotaxi service.
@@route55qatar According to my information, Tesla was marketing their tech as "FSD" too aggressively, and after a lethal accident MobileEye decided to split.
Such impressive progress by Mobileye! Also, a great pragmatic pitch on the event horizon / timeline to a L2+ world. Make sure to listen to his summary takeaways at the end of the presentation - insightful, exciting, and realistic.
Dear Amnon/Mobileye, you and your team are hero to the industry/society and a huge inspiration..... can Mobileye someday publish which OEM cars use the most Mobileye technology that people can buy? Safety is so important, lots of people are making car buying decisions based on the ADAS/AV features and the technology behind it. We would really like to know which car we buy will be using the most Mobileye technology!!!
@@boli2940 Developer showcases are nice and all. But I don't think you can really judge how good it is compared to Tesla without having users to test them thoroughly. I work in the software industry and in my experience you can test internally all you want, the most bugs and system design flaws you only find out after giving the public access to it. 1 week of public testing is worth more than years of internal testing, just because of the sheer amount of people testing and trying different things.
What I would like to see is left turn without guiding lines on the streets, as they have in the video. I have never seen those in real life. and I would say it helps immensely to guide the vehicle. great progress nevertheless.
Driving without guiding lines is NOT a very big issue since the sensors recognize road shape, traffic direction, object movements, road signs, tree, pedestrians, dog, cat and even Mcdonald signs. In fact it is more safer than human since we only have two eyes not twelve. Ps. You will be in trouble when driving in the night with heavy smog in front of you. Sensors have no problem since it cna recognise road shape and other shape.
Johan van Dalen I do understand that. What I meant is that in most countries I visited, there is now markings inside intersections. So the question was, how will it handle?
@@matejnimmerrichter768 How do u YOU handle your car in that situation? Ps. I repeat : Driving without guiding lines is not a very big issue since the sensors recognize road shape, traffic direction, object movements, road signs, tree, pedestrians, dog, cat and even Mcdonald signs.
@@ashh3051 The difference is in the cloud-storage technology. Tesla will have to solve everything in realtime, while Mobileye, Waymo, GM etc will get help from the cloud. So signs, potholes, road-layout etc is already known before the car arrives, which makes the task way easier to solve.
@@oophus1 I guess you meant that Mobileye et al use online High Definition maps. Yes they are incredibly helpful, but I can imagine Tesla would be using one as well. Any AV maker with sufficiently powerful and accurate perception system can make and use HD maps.
@@user-to2rf1rj5v Of course, they can't be solely relied on. In fact to be able to use the map information, you need to have a good vision system that detects the positions of static infrastructures which is then used to precisely localize on the map, apart from the GPS inputs. They are incredibly useful for things like automatic lane changes, automatically staying within speed limits and anticipating exits way ahead. They are instrumental for L3/4+ autonomy. Speaking from the experience working as a HD map engineer. 🤓
Tesla's Andrej Karpathy has confirmed in a presentation that Tesla also builds maps as they drive with stop signs, potholes etc. But I think their goal is that the vision system won't need the map input anymore some day
Honestly it seems Tesla copied mobile eye, Tesla originally used mobile eye in it's first cars, then they went on to develop their own system which is oddly similar... Although Tesla is very passionate about NOT using HD mapping... I wish both companies success, (although Tesla has WAYYYY more data to play with) 😀
They have a massively different approach to the problem. Inspiration in both ways. Also, Tesla was the proponent of visual-only from the start. Even with Mobile Eye tech.
Tesla does not have more data to play ; MBY has millions of cars on the road, not only one brand. To map the world real time you need more than 500K Teslas a year. To have a good coverage you need to be in GMs, VWs, Toyota's, the volume cars. Tesla needs to partner with others to get coverage or have potentially a less safer car.
Tesla Tube you have read the fine print. He told HD map base on Lidar , where you need GB of data. I doubt they do not use any he map with sparse landmarks. At least the announcement they made in 2016 was saying different.
If I understand correctly they use road data from other manufacturers to create their own maps for locomotion. Do those cars all have 360° video feeds? How are they going to tackle the thousands of weird edge cases without public user testing and without access to user video upload functionality? Don't they have to remap everything constantly to have accurate maps (because of construction/road changes etc.)? Nonetheless great to see there is healthy competition to Tesla in the autonomy space. Autonomy can't happen soon enough, it will save so many lifes. Humans are horrible drivers.
they create their own hd maps by crowdsourcing the data from the millions of cars that already are equipped with a front-facing mobileye camera. These cars just send very little amount of data (10kb/km) from which the maps are created and updated automatically. I'd also like to know, how they source video for AI training. Does anyone know?
@@sadburai1 That's the thing, you can collect road data for HD maps with those cars, but not video footage for training a neural net. The neural net is in the end what decides on how the car behaves on the road, basically the brain of the autopilot, HD maps only get used for where you are positioned on the road.
A long journey starts with the first step, and then the second. Eventually we will get there. The timelines are goals for individuals and companies as a guideline and force behind schedules. When solving a complex problem, you MUST divide it into reasonably sized problems and solve each individual problem on its own and them assemble the whole complex solution. Imagine how they went to the moon and back BEFORE the end of 70's in the previous century. That was more than 50 years ago. I am optimistic that the self driving car paradigm will be solved in stages (as professor Shashua is showing in one of the slides) and in timely manner and the whole humanity will benefit from it.
Ali from Home are but Elon talks so much bs, and honestly anyone who has to lie to start any journey is crappy human. Tesla has soured me to EV’s and FSD.
The system will not work outside of cities. What about fog, snow, animals near roads, icy roads, unpaved roads. A city-summer-sunshine-mode would be an expensive addon for a car.
Don't confuse your eyes with sensors which come is many forms. Let me put this way : A missile can find a small car on the ground from 3 km height AT NIGHT (!) near Bagdad airport recently. Even if there was thick smog in the air the missile will find the car easily. Do you think car technology can't do the same?
@@gsopoagmle Bagdad? The city where Aladdin met Jasmine and took her to a tour of the world on his "autonomous driving" carpet Bagdad? Ok. In the case of the Bagdad missile, there might have been laser painting of the target! But in the case of autonomous driving, a combination of cameras, sensors and other sensory devices can "fully" "see" all that is needed to "see" in order to autonomously drive. THAT is the salient point.
@@moeali12345 Second generation lidar can see everything even in bad weather condition. It take a small effort for eye mobile to implement Ladar in their package.
@@gsopoagmle Yeah but that's the thing, they don't use lidar or radar anymore. They use camera feed only, same as Tesla nowdays. I cant talk for Mobileye, but Tesla drives without issues in heavy fog, according to the visualisation on the screen the car sees more than a human, despite camera only. And yes it also detects ice or puddles. I think Mobileye is competent enough to also solve this.
As a Tesla shareholder, I am surprised to see a competitor that somehow inspires me more confidence in what they are doing. Precise information rather than overwhelming non- backed up promises. My prediction would be that while MobileEye probably will have an edge in full autonomous driving, the vertical integration will Tesla allow to compete successful. That is, if both companies self driving systems will be authorized. If only one makes it onto the streets, that company will dramatically win market share. In the case of MobilEye this means a number of OEMs plus Nio plus their own Robotaxi service.
Mobileye was Tesla's ADAS partner in the early days. Disagreement on core principles parted their way.
@@route55qatar According to my information, Tesla was marketing their tech as "FSD" too aggressively, and after a lethal accident MobileEye decided to split.
Such impressive progress by Mobileye! Also, a great pragmatic pitch on the event horizon / timeline to a L2+ world. Make sure to listen to his summary takeaways at the end of the presentation - insightful, exciting, and realistic.
Dear Amnon/Mobileye, you and your team are hero to the industry/society and a huge inspiration..... can Mobileye someday publish which OEM cars use the most Mobileye technology that people can buy? Safety is so important, lots of people are making car buying decisions based on the ADAS/AV features and the technology behind it. We would really like to know which car we buy will be using the most Mobileye technology!!!
This Guy is awesome!
Genius!
Why Intel stock hasn’t appreciated baffles me.
It took me a while to really it the letter "i" (intel) stand in front of the letter "m" (mobileye) in the post-acquisition logo.
Just rewatched Tesla's autonomy day and very interesting to compare both approaches, both got rid of lidar for example (edit: apparently not entirely)
I'm long on intel for this mostly
1. Awesome 2. Seems further then Weimo in general 3. In some aspects further then Tesla, in others not 4. 2+ sounds like Tesla's solutions (no Lidar)
David Rubli this is way more advanced than Tesla...
David Rubli Tesla says no lidar for level 5, so you’re wrong
@@boli2940 Developer showcases are nice and all. But I don't think you can really judge how good it is compared to Tesla without having users to test them thoroughly. I work in the software industry and in my experience you can test internally all you want, the most bugs and system design flaws you only find out after giving the public access to it. 1 week of public testing is worth more than years of internal testing, just because of the sheer amount of people testing and trying different things.
I had an ad about this.
wtf!! You have shown wrong map of INDIA (40:19) . You haven't included KASHMIR !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
What I would like to see is left turn without guiding lines on the streets, as they have in the video. I have never seen those in real life. and I would say it helps immensely to guide the vehicle. great progress nevertheless.
Driving without guiding lines is NOT a very big issue since the sensors recognize road shape, traffic direction, object movements, road signs, tree, pedestrians, dog, cat and even Mcdonald signs.
In fact it is more safer than human since we only have two eyes not twelve.
Ps.
You will be in trouble when driving in the night with heavy smog in front of you. Sensors have no problem since it cna recognise road shape and other shape.
Johan van Dalen I do understand that. What I meant is that in most countries I visited, there is now markings inside intersections. So the question was, how will it handle?
@@matejnimmerrichter768
How do u YOU handle your car in that situation?
Ps.
I repeat :
Driving without guiding lines is not a very big issue since the sensors recognize road shape, traffic direction, object movements, road signs, tree, pedestrians, dog, cat and even Mcdonald signs.
Johan van Dalen aahhw, sorry a misread your comment. I overlooked NOT :)
I wonder how this compares to Tesla's vision-only efforts?
Tesla is not vision-only currently. They use radar as well.
@@ashh3051 The difference is in the cloud-storage technology. Tesla will have to solve everything in realtime, while Mobileye, Waymo, GM etc will get help from the cloud. So signs, potholes, road-layout etc is already known before the car arrives, which makes the task way easier to solve.
@@oophus1 I guess you meant that Mobileye et al use online High Definition maps. Yes they are incredibly helpful, but I can imagine Tesla would be using one as well. Any AV maker with sufficiently powerful and accurate perception system can make and use HD maps.
@@user-to2rf1rj5v Of course, they can't be solely relied on. In fact to be able to use the map information, you need to have a good vision system that detects the positions of static infrastructures which is then used to precisely localize on the map, apart from the GPS inputs. They are incredibly useful for things like automatic lane changes, automatically staying within speed limits and anticipating exits way ahead. They are instrumental for L3/4+ autonomy. Speaking from the experience working as a HD map engineer. 🤓
Tesla's Andrej Karpathy has confirmed in a presentation that Tesla also builds maps as they drive with stop signs, potholes etc. But I think their goal is that the vision system won't need the map input anymore some day
Honestly it seems Tesla copied mobile eye, Tesla originally used mobile eye in it's first cars, then they went on to develop their own system which is oddly similar... Although Tesla is very passionate about NOT using HD mapping... I wish both companies success, (although Tesla has WAYYYY more data to play with) 😀
They have a massively different approach to the problem. Inspiration in both ways. Also, Tesla was the proponent of visual-only from the start. Even with Mobile Eye tech.
Tesla does not have more data to play ; MBY has millions of cars on the road, not only one brand. To map the world real time you need more than 500K Teslas a year. To have a good coverage you need to be in GMs, VWs, Toyota's, the volume cars. Tesla needs to partner with others to get coverage or have potentially a less safer car.
Telsa has the approach of not needing to map the world... Just train the NN and then apply it to anywhere in the world...
Tesla Tube you have read the fine print. He told HD map base on Lidar , where you need GB of data. I doubt they do not use any he map with sparse landmarks. At least the announcement they made in 2016 was saying different.
@@blanamaxima Tesla is very clear that HD maps is not the way. They also have more footage that is relevant of teaching the NN than Mobileeye.
Am I the only one thinking to myself if Tesla or Elon is watching is?
Super cool project
I heard Tesla uses the Thomas Guide 😂
very interesting............
90mm^2 EQ5 is not bad at all
If I understand correctly they use road data from other manufacturers to create their own maps for locomotion. Do those cars all have 360° video feeds? How are they going to tackle the thousands of weird edge cases without public user testing and without access to user video upload functionality? Don't they have to remap everything constantly to have accurate maps (because of construction/road changes etc.)?
Nonetheless great to see there is healthy competition to Tesla in the autonomy space. Autonomy can't happen soon enough, it will save so many lifes. Humans are horrible drivers.
they create their own hd maps by crowdsourcing the data from the millions of cars that already are equipped with a front-facing mobileye camera. These cars just send very little amount of data (10kb/km) from which the maps are created and updated automatically. I'd also like to know, how they source video for AI training. Does anyone know?
@@sadburai1 That's the thing, you can collect road data for HD maps with those cars, but not video footage for training a neural net.
The neural net is in the end what decides on how the car behaves on the road, basically the brain of the autopilot, HD maps only get used for where you are positioned on the road.
A lot of snow and ice in Jerusalem?
A long journey starts with the first step, and then the second. Eventually we will get there. The timelines are goals for individuals and companies as a guideline and force behind schedules. When solving a complex problem, you MUST divide it into reasonably sized problems and solve each individual problem on its own and them assemble the whole complex solution. Imagine how they went to the moon and back BEFORE the end of 70's in the previous century. That was more than 50 years ago. I am optimistic that the self driving car paradigm will be solved in stages (as professor Shashua is showing in one of the slides) and in timely manner and the whole humanity will benefit from it.
Ali from Home are but Elon talks so much bs, and honestly anyone who has to lie to start any journey is crappy human. Tesla has soured me to EV’s and FSD.
The system will not work outside of cities. What about fog, snow, animals near roads, icy roads, unpaved roads. A city-summer-sunshine-mode would be an expensive addon for a car.
Don't confuse your eyes with sensors which come is many forms.
Let me put this way :
A missile can find a small car on the ground from 3 km height AT NIGHT (!) near Bagdad airport recently. Even if there was thick smog in the air the missile will find the car easily.
Do you think car technology can't do the same?
@@gsopoagmle Bagdad? The city where Aladdin met Jasmine and took her to a tour of the world on his "autonomous driving" carpet Bagdad? Ok. In the case of the Bagdad missile, there might have been laser painting of the target! But in the case of autonomous driving, a combination of cameras, sensors and other sensory devices can "fully" "see" all that is needed to "see" in order to autonomously drive. THAT is the salient point.
@@moeali12345
Second generation lidar can see everything even in bad weather condition.
It take a small effort for eye mobile to implement Ladar in their package.
@@gsopoagmle I get it. I agree with you. MobilEye is a good company as is Intel. They will make this work.
@@gsopoagmle Yeah but that's the thing, they don't use lidar or radar anymore. They use camera feed only, same as Tesla nowdays. I cant talk for Mobileye, but Tesla drives without issues in heavy fog, according to the visualisation on the screen the car sees more than a human, despite camera only. And yes it also detects ice or puddles. I think Mobileye is competent enough to also solve this.