Unedited Ride in Mobileye’s Self-Driving Car

Mobileye is developing for the self-driving car two truly redundant sensing systems: one with surround-view cameras alone, and the other with radars and lidars. In this unedited demonstration of the camera-only technology, you can see Mobileye’s autonomous car successfully navigating a complex driving environment rich with pedestrians, unguarded intersections, delivery vehicles and more.
At certain points of interest during the ride, we display additional sensing features that are not shown in the unified 3D environmental model (for the sake of reducing clutter). This under-the-hood view demonstrates the depth of our highly sophisticated surround sensing and RSS-enabled Driving Policy.
An Hour with Amnon Autonomous Car Powered by Mobileye CES 2020 - • CES 2020: An Hour with...
📈 Watch the first Mobileye Investor Summit - • 📈 Watch the first Mobi...
Prof. Amnon Shashua: The Promise of Machine Learning and AI in Transforming Industries - • Prof. Amnon Shashua: T...
Responsibility-Sensitive Safety Explained by Jack Weast II Mobileye - • Responsibility-Sensiti...
Prof. Amnon Shashua Gives Keynote Address at AI Week 2019 in Tel Aviv - • Prof. Amnon Shashua Gi...
RSS - a Mathematical Model to Guarantee AV Safety - • RSS a Mathematical Mod...
Prof. Shashua at World Knowledge Forum: Platform for Safe & Scalable AVs • Prof. Shashua at World...
Three Major Pillars of Technology for Autonomous Car CES 2017 - • Autonomous Cars Three ...
Subscribe now to Mobileye on KZbin: bit.ly/2vRes7k
About Mobileye:
Mobileye's advanced driver assistance systems (ADAS) technology is deployed in more than 50 million vehicles today and is integrated into hundreds of new car models from the world's major automakers including Audi, BMW, FCA, Ford, General Motors, Honda, Hyundai, Kia, Nissan, Volkswagen, and more. Mobileye began with the vision of reducing vehicle collisions and resulting injuries and fatalities. Today, Mobileye makes one of the most advanced collision avoidance systems on the market, while working toward autonomous driving and the coming autonomous mobility-as-a-service (MaaS) revolution in road safety.
 
Connect with Mobileye:
Visit the Mobileye WEBSITE: www.mobileye.com/us/fleets/
Like Mobileye on FACEBOOK: / mobileye
Follow Mobileye on TWITTER: / mobileye
Join Mobileye on LinkedIn: / mobileye
Mobileye® 8 Connect™ - Driven by Safety: • Mobileye® 8 Connect™ -...

[@Vikktor97]

Most impressive self-driving situations in this video: 1. Unprotected left turn 2. Unprotected right turn 3. Into a roundabout 4. Deal with a roadblock on the only lane.

[@liniumz]

The car company that's implementing your technology will get all my car budget from now on. I want full self driving car more than anything.

[@paulbyatt3219]

I thought driverless cars would be great for the disabled & the elderly, but it seems the younger generation really wants them as they want everything done for them.

[@petchlnwzaaa]

J L Yeah. The city where I lived as really bad traffic jam. So it’d be cool if I can have self-driving car will I sleep while go to work 😂

[@liniumz]

@@paulbyatt3219 wow grandpa has something against the younger generation... I am sure you take zero advantage of the technology, prob living in a cave to stay fully independent...

[@Glenhh]

@@liniumz I guess he is also flying his commercial airplane himself like old people do you know?

[@connelly6375]

this system is so much better than the tesla FSD videos i've been seeing

[@yuanfang2352]

the most impressive thing for me is to run 12 cameras on two 7nm chips. Mobileye has really pushed the edges for the industry.

[@1shaymark]

Good job, the full self driving is comming

[@peterpoulsen4794]

Cool tech, but if i was driving behind it i would wonder if it was an old grandmother driving the car ;)

[@michelgent7419]

I heard the auto pilot is based on one of the developer's grandmother driving

[@AudiTTQuattro2003]

My thoughts exactly. Too cautious sometimes (given what a competent driver would do) and used edging out into traffic instead of anticipating openings. Truely impressive but I would hate a whole city of these things.

[@MarkoTintor]

Very impressive! Scene display looks a little delayed compared to drone footage.

[@zhenqiao70]

Does it rely on RTK GPS to do licalization? Or only use normal GPS with camera lane detection and multiple sensor fusion?

[@trott2006]

quite impressive. regarding "grandma" driving style this is probably due to all the not-so smart cars / drivers. The switch from anticipation of other drivers behaviour to knowing other cars behaviour by connected communication will be the game changer

[@ButterNutter37]

This is absolutely magnificent! Bravo!

[@deltajohnny]

Awesome. . 👏👏👏👏👏👏 Thank you for your work, and I wonder when will this technology will be available? I can't wait!!! 😉😉😉👍👍👍

[@boli2940]

Delta Johnny the question is not when is this available, the question is which car company will be using it. Tesla, GM, Ford, Volvo, Toyota, and Mercedes, etc are all too afraid to be dependent on Mobileye, therefore decided to use their own less reliable, less safe, less advanced in-house systems. These stupid car makers are slowing down our advance to self driving cars, because Mobileye is light years ahead of everyone else. Most advanced and most aligned with Mobileye now is BMW, Nio, Lucid, Nissan to some degree. In my opinion, only buy cars that are fully based on Mobileye technology, AV and safety is too important. Don’t be a victim and become the other car manufacturers’ lab rats.

[@russm8193]

@@boli2940 They are primarily a hw company. How do their Neural Net sw engineers compare w Waymo and Apple?

[@Glenhh]

@@boli2940 BMW hahaha, what are you talking about. Show me a BMW that drives trough a city like that.

[@whateverrevetahw6846]

Zeekr 001 late this year will be the first car released with their SuperVision system but only in at first China.

[@dalongli6624]

at 3:42, what are those objects indicated with red color? it is at the left of the vehicle, at the intersection where there was a left turn sign on the ground

[@aptyp813]

no stop zone

[@okazakijap]

Fantastic.

[@nkallen123]

Super cool!

[@peterb.5845]

Erstklassig Top Wahnsinn 👍🏻😉😁🤗

[@bobvanwagner6099]

Can you add in the speed display, also something showing the gear and brake force applied? Something like a vectored accelerameter display? Also there seems to be difficulty tracking scooters.

[@gusorviston1]

speed is in the upper left video square

[@bobvanwagner6099]

@@gusorviston1 thanks, but i meant the gear and brake force might be included on that.

[@eugenealooeff2824]

Why it turns wheels when car stays?

[@mikemorrell7921]

Very impressive. Do you know why the vehicle stopped at about 10:42, 11:30, and 12:43? Did it think the pedestrians it saw were going to cross, as indicated by red people? The yielding to the occluded person was really impressive. The camera didnt appear to catch them until they passed the car so is it more likely the stop was because of the slow down of the car in front and the car coming the other way coming into lane? Why the stop near 12:43? It thought the parked car was coming to a stop? Was the algorithm ready to commit suicide at 14:35 :) The dancing rubbish bin(?) is cute at 16:00. I think the algorithm had signed is resignation letter by the bicycle at 17:02. Again, very good work and thanks for sharing.

[@sotongbaby1]

5:08 There's a mistake that the car can't detect the white line well

[@Nick-li1zw]

Hello does the mobile eye record in the event of a wreck to prove you were or weren't at fault?

[@pelatihfinansial]

there should be, it is fairly easy to prove whether there is driver interference or not.

[@JD-kf2ki]

I wish I could work for this company.

[@angelsancheese]

Thank you for the video! It's crazy that mobileye does not have more than 300,000 cars collecting data, yet they can still make a self driving system in Jerusalem. This demonstration is very impressive and I'm glad it only used cameras, yet I think it was done by having a downloaded mapped version of the environment in Jerusalem. It would be nice to know if it is using a downloaded version of the environment.. ? Thank you!

[@dariopavlovic8640]

Angel Sanchez from what I read, mobil eye is installed in some form into 17 million cars these days.

[@angelsancheese]

Dario Pavlovic yes but they do not have more than 300,000 cars collecting data and sending that data back to mobileye

[@dariopavlovic8640]

Angel Sanchez i don’t know anything about MobilEye details, but last year they signed a deal to provide their system into 8 million cars. I highly doubt only 300k would transmit data back to mobileye, but could be possible. In any case, their demo here is quite impressive, although far from perfect. Better than Tesla’s “fsd” from what I’ve observed while driving a Tesla.

[@angelsancheese]

Dario Pavlovic thank you

[@HardestTorkum]

@@dariopavlovic8640 the only difference is that you can try out a Tesla anytime on any road you want. This demonstration only works in Jerusalem and who knows what the MobilEye engineers have hard-coded here in order to drive this specific route.

[@devashishraval7811]

Excellent technology..... all smartness lies in two small chips.... a mild music would make this video more interesting....

[@pawankumarhegde9296]

Why is there a big flicker in the drone camera at 10:09? and immediately there is an explanation of object detection for yielding scenario instead of instance segmentation ?The Road construction was superbly handled . Hmm !

[@manliobarajas2318]

Is there a demo like this... but at night or snowy location?

[@izybit]

Yandex had a demo on snowy Moscow streets. Day/night won't make much of a difference as long as you have headlights on. Lot's of Tesla videos on youtube show their system performing really well in night/snow/rain. Mobileye should be able to do the same.

[@pccarraway9632]

2017 Mobieye Patent: "Building night vision and other driver assistance systems (DAS) using near infra-red (NIR) illumination and a rolling shutter" covers some of your question patents.justia.com/patent/10129465

[@stri8ted]

@@izybit Would be very interesting to see how MobileEye compares with Tesla. It seems that Tesla uses more of a "black box" approach.

[@SuperCheess]

was there any use of high resolution maps in advance? Otherwise, really good, ahead of competition!

[@russm8193]

They are dependent on HD maps. It's their whole biz model.

[@michelgent7419]

Every production AV car will have HD maps. No regulator will let AV car drive without one.

[@russm8193]

@@michelgent7419 Proof is in # of miles driven per intervention by driver, and # of accidents.

[@sddndsiduae4b-688]

@@michelgent7419 Tesla tries to do without it, but for now it all looks really bad, i.e. when they initially enabled navigate on autopilot in AU, it did missed correct highway out.

[@sddndsiduae4b-688]

@@russm8193 miles on highway (navigate on autopilot) is actually nothing compared to miles inside city with pedestrians around.

[@vladmihai306]

This is a real life GTA 2 :))

[@mgore7624]

Impressive how many people breaking road rules, drive over double lines, drive in the lane of others,.. The drone could used to connect with the mobileye to scan the lane behind the next car in front. And it's not clear why the car is drive into the roundabout when a car comes around? 9:50 and 11:30

[@guanyu210379]

Recognition of objects still needs improvement but during good recognition, the position estimation looks very good. As for the localization part, that looked like HD-Map combined with dGPS to me. Not 100% sure tho'. The driving function parts look awesome! The car runs quite stable.

[@Joraultechy]

Yu Guan definitely HD Maps. Object recognition isn’t strong. I’m pretty surprised at the rendering, it seems mainly informed by HD Maps and not the NN.

[@sx725]

The CEO mentioned they rely on HD maps in their CES keynote.

[@Alacaalam]

We live in Toronto. I have 3 cameras on my car. All of them so dirty after 5 minutes of drive that I can barely see anything on the screen... How you can drive automatically using such dirty cameras?

[@leibmark]

I hope that the cameras will clean themselves like an eye.

[@stri8ted]

Interesting point. Presumably it can utilize similar mechanisms that are used to clear your windows.

[@TomBertalan]

@@leibmark An eye on an automobile? Like, some sort of mobile-eye?? That's crazy talk

[@leibmark]

@@TomBertalan я говорил о системе очистки камеры. Это трудно понять?

[@NSN2195]

Perfect. We need a similar video for a dark snowy night

[@JohannesSchmitz]

Human drivers drive better in areas they have explored before. There's a type of map stored in our brain. I wonder why so many people want the car to not use map data although it is available. Also stuff like temporary road blocks and construction work can be detected by the car and overlayed on top of the static map.

[@dospesentas]

Many comments seem to think this is a production vehicle and should be perfect and function in extreme conditions. This is a prototype and given the many technologies involved is pretty amazing. Five years from now this will be nothing special.

[@russm8193]

Compare with Waymo.

[@sam-zg5ke]

great thing it has street manoeuvring hope to see it on hard roads with less instructions

[@gabrielswsia3369]

למה אין בעיברית

[@roi12555]

הכניסה לכיכר ב9:50 נראת לי קצת מסוכנת

[@Scott-sm9nm]

The yellow triangle icons/emblems over the cars to 'watch' for or be cautious about was very interesting. i.e. anticipate potential problems.

[@McGybrush]

Nicht schlecht. Aber ich glaube ich würde erst Jubeln wenn die Strasse nicht schon im System hinterlegt ist. Diese wird leider nicht in Realtime gesichtet. Somit für eine Baustelle und Witterungsbedingte Änderungen an der Fahrspur noch nicht geeignet. Also alles was das System noch nicht weiss. Aber dennoch gut wie das System auf Fussgänger und ähnliche Variablen reagiert. Nur Minimal defensiver als ich selbst fahren würde.

[@alokcom]

no LIDAR?

[@FinalHopee]

Alok Prasad As Musk said, once the vision will be fully resolved there will be no need for expensive LiDAR.

[@knp4356]

Camera only works in places with no-rain and snow. Introduce a little challenging environment and the system becomes very unreliable. You guys need to integrate radar/lidar into this architecture to create a turn-key solution for car OEMs. Great stuff here. Thanks.

[@arielgalil]

מדהים. אני מתחיל להאמין שמכוניות אוטונומיות אפשריות בעתיד הנראה לעין.

[@roi12555]

להההההההההה

[@MaxCaud]

Was this area pre-scanned? Or was everything preformed in real-time? Very cool.

[@izybit]

They use HD maps so road markings, signs, etc come from the maps, the car just understands where it's at and acts accordingly.

[@yuanfang2352]

the next thing is to cope with all kinds of weather and light conditions...

[@raul_pesch]

Was the environment mapped before driving or is this real time mapping?

[@udishomer5852]

Its mapped. Mobileye has an automatic HD mapping system powered by 1 million cars with adas systems. US, Europe and Japan will be fully mapped by end of 2020.

[@joekkl]

Map is static data. That means there's a risk it will get outdated.

[@udishomer5852]

@@joekkl Its not static, its constantly updated by cars with installed Mobileye tech. Mobileye expects 14 million such cars by end of 2022.

[@joekkl]

@@udishomer5852 You said: "US, Europe and Japan will be fully mapped by end of 2020." That means the map info is not real time. Imagine a construction site re-route the road, how will the car knows the latest road condition without constant map update? If the car cannot recognize roads in real time and require a pre-defined map as a guide, then it's not smart enough to take over from the driver.

[@udishomer5852]

@@joekkl again, its mapped by cars driven by regular people. Every car that drives with mobileye tech and the brand has an agreement with Mobileye, is a mapping device. There will be 14 million such mapping devices by end of 2022. As for how depended the cars are on the HD maps, I don't know, it was not discussed in the recent presentations.

[@bogao7506]

地图是预先绘制好的高精地图对吗

[@gusorviston1]

Yes.. .errr No.

[@tatianashkvyra7065]

❤❤❤

[@danfrederiksen1607]

I assume it uses GPS of course. Just +-1m precision or RTK?

[@sx725]

Probably rtk

[@danfrederiksen1607]

@@sx725 RTK is borderline cheating. And you probably can't rely on that in all countries and it doesn't work among tall buildings afaik. I figure they are smart enough to do it without RTK. They also 'cheat' by having HD road maps so they can position by that.

[@blueblimp]

At 21:48, it seems to detect the oncoming bike quite late.

[@sddndsiduae4b-688]

well its not perfect that's why they said they need lidar to drive without safety driver, but if you compare it to kzbin.info/www/bejne/hHLFhXSnrdd_oKc you would obviously see difference.

[@stri8ted]

Not sure its considered "late" relative to computer reaction times. We also don't know what the scene looks like from a human POV, to compare effectiveness.

[@sddndsiduae4b-688]

@@stri8ted late means human would detect it much early - enable/download 4k video (you can download video part via network monitoring in browser) and concentrate on left down screen part, bike is visible for human for long time before it was detected by car.

[@sddndsiduae4b-688]

@@stri8ted bike is clearly visible to human driver when it drives out of occlusion of white car (which is parked right after parked blue car) i.e. even early than it visible via drone.

[@roi12555]

that's impressive

[@juniorjose06]

City skyline is looking good on this video.

[@CookiePepper]

Impressive if it can do on a road where the car has never driven before.

[@ShopTown]

I would keep my fusion if you will sell me this 😭

[@gsms5846]

While watching this I realized it's not hard tp make an autonomous vehice.

[@GM-zn3pt]

great stuff. 4.29 lots of false positive pedestrians

[@IsraeliXdude]

Better false positive than false negative :)

[@GM-zn3pt]

@@IsraeliXdude 100% right. Just still needs to have a high availability or else nobody will actually get from A to B on time

[@Root6532]

Impressive work! Wow. But, when it starts to rain/snow/fog this won't work. If you can't see something neither the camera.

[@TomBertalan]

Are you, as a human, able to drive in such conditions?

[@Crawfish411]

@@TomBertalan yes

[@user-RCST]

@@TomBertalan a human is not a computer.

[@D00M475a]

@@user-RCST Does a human rely on vision to drive?

[@user-RCST]

@@D00M475a yes

[@ibsn87]

Looks like high precision GPS and maps is doing the majority of the work.... and the cameras are identifying threats and positioning - very nice. This obviously works until you get GPS errors or database map errors Am I right? This requires high precision maps - something Tesla wanted to avoid. If a GPS fails or the database is incorrect there’s a problem. Tesla has the neural network, machine learning and MOST importantly millions of cars on the road (fleet) to continue learning from. Tesla determines its surroundings on the fly as opposed to being told by preloaded maps. Looks like they are using a combination of machine vision and maps which will work very well, but the problem is, Someone has to pay for these VERY high precision maps that will require constant upkeep and maintenance!!! Impressive nonetheless the car and the team did an amazing job to autonomously navigate these complex roads!!! My question is, how many thousands of hours have gone into this planned route. - and who pays for the map construction and maintenance of the world!? $$$$ cost is everything.

[@zhenjilu7125]

I am not convinced this is a camera only demo, if as claimed, the car would follow the bus to the bus lane around 5:00.

[@russm8193]

Why?

[@NickLukefahr]

Lanes are clearly marked, the car would have had to exit its lane to follow the bus.

[@zhenjilu7125]

@@NickLukefahr true, but a bus lane marking on the left was also identified.

[@texastriguy]

This uses highly accurate pre-mapping of all streets for the drive, correct?

[@jamesyoung8797]

This bodes well for Tesla because it demonstrates Tesla's camera based FSD system should work equally well. There is no reason why Tesla is not already at a similar level. The only difference is, once available, hundreds of thousands of Tesla vehicles will have the capabilities overnight. I guess there is a good reason Tesla is setting up a shop in Israel, keep your competitor close.🤣 Who knows, a renewed Tesla and Mobileye partnership would not surprise me. FSD needs all the help it can get right now.

[@sddndsiduae4b-688]

nope kzbin.info/www/bejne/hHLFhXSnrdd_oKc it actually explains why tesla official demo of FSD don't have ANY people around car. Tesla is far behind...

[@shahbazfawbush]

Cooperation for the win!

[@mendi1122]

In 3:44 the car should at least slow down.

[@mendi1122]

When the driver is not aware to his limitations it is very high risk driving.

[@binsun8384]

If any automaker use this system for their self driving feature, and there is accident happened, who's going to take the responsible. The system provider or car company?

[@lidmon]

Yeah right! They're using detail road map, with lanes and pedestrian crossings. And even that map doesn't cover every street. So it's easy for autopilot but totally unuseable in real life. And this video rises a lot of questions. One car was making three way u turn, and the AP supposedly saw blinker, but miss them on every other turning car. Also stop lights of vehicles in front. AP tried several times to bypass car jams. Several times traffic light was visible although car was far away from intersection. Totally unrecognisable surroundings. AP didn't recognize upcoming traffic when there was a dividing island on road.

[@michelgent7419]

1. That's how people drive in Israel 😅 2. Part of their business model is to sell HD maps of the roads of the entire world -> so I guess they'll have the maps they need

[@Glenhh]

@@michelgent7419 So they map the entire world every single day? I don´t want to rely on a map that is one month or one year old. Construction and other things could change the environment completely.

[@michelgent7419]

@@Glenhh A great point. And how do you think it could be solved?

[@Glenhh]

@@michelgent7419 Don´t rely on maps, drive like a human does. That´s pretty hard but vision has to be the decider in every situation. That could mean that we need longer to get to autonomous driving, but at least it will be really autonomous not like the Waymo cars...

[@michelgent7419]

@@Glenhh AV has to be much safer than human drivers. We as a society can accept death from car accidents from human driving. The tolerance for AV accidents is much much lower. Map is essential. So how would you solve it? :)

[@M4XC4V413R4]

Where is this? The amount of broken road rules one can see in this singles video (performed by human drivers) is too damn high. Also, the city needs to learn how to make intersections for the love of god. It's not like there isn't enough examples in the world.

[@tadangrai5256]

Not convinced

[@Digalog]

still pretty impressive x)

[@raashidbhatt1]

GTA -2

[@phatfel]

First?... Anyways very impressive indeed!

[@Ainappa]

2:02 lol! That's a pretty bad laggy Oversegmentation of a pedestrian. Overall seems to be pretty laggy with bad yaws of cars. This Demo view lacks real time info that usually gets fed into the monitor.

[@vladhristov2316]

Markings are clear, all the environment is well under sunlight. Come here where I am and there is nothing on the street. Some of the signs are paint sprayed, street marking are vague and the road is full with patholes. It is anarchy.

[@avi123]

6:10 the car turns into a bus

[@gusorviston1]

uh.. no.. it is stopped for a merge and bus comes up in the lane beside it. Then the car allows the bus to go since it is ahead of it. Car never moves when the bus is near it until the bus leaves. take ur glasses off and rewatch it.

[@avi123]

@@gusorviston1 what I mean is on the blue screen where it shows the systems perception of the situation, it shows a really long car, then it realizes it is actually a bus.

[@denniscraig1708]

Ah. Got it

[@richman9647]

💖🇮🇱🕎✡️🇮🇱💖

[@awaken69]

get back to us when u dont need HDmaps anymore..

[@alph4966]

✡🇮🇱

[@nonconsensualopinion]

Camera based autonomous system with a huge vertical center screen in the cabin? Nice Tesla you built there! Seriously though, great work. Great autonomous performance.

[@michelgent7419]

Mobileye was the one who built Tesla's system..

[@udishomer5852]

The screen is for the testers, not for the customers. And Mobileye is probably way ahead of Tesla in complex city driving environments.

[@michelgent7419]

Mobileye sells to car companies, Tesla to customers. Thats why the general public might think Tesla system was first. Mobileye's system is way a head of Tesla's

[@azizsaba2379]

screw tesla

[@brianraynor5559]

@@michelgent7419 Initially true (in v1 of AutoPilot was indeed MobileEye). v2 and 2.5 was NVidia based (Tesla software), v3 is Tesla HW and Tesla SW. This is awesome though - definitely shows how good MobileEye's system is. Few minor visual misidentifications (a few times it missed people, thought a shadow was a person, etc), but overall looks like even the visualization is 95% or better. Very good. Teslas with v3 HW are getting close, visulations in what is released into the wild is on par overall with what this video showed for MobileEye - Telsa appears to see more "types" of objects (pylons, trashcans, road markings (right turn arrows, etc)), but from this video seems MobileEye shows a bit more detail on what it does indentify and can see a bit farther. Of course this is all based on the prerelease MobileEye vs. the released Tesla FSD features in v3 (not currently fully full self driving). So imagine Tesla is a lot closer to MobileEye than what is shown (internally that is vs what they have released in the wild). That said, this video from MobileEye is downright impressive!! My Model 3 is currently on v2.5 hardware, so I don't get some of the newer features and visualizations, though I bought FSD, so I'm on the list for the free upgrade to v3. So basing off videos and reports of v3. Hopefully I'll get it soon. MobileEye has definitely confirmed and proved that camera only systems can do an awesome job without Lidar.

[@RobertoPradoF]

Impressive demo, but I think Tesla will get far away from the competition once the Starlink's satellite constellation is working. At that time, all Tesla cars will drive autonomously and safely in any part of the world.

[@frednicholson]

Why woukd starlink provide any benefits to self driving? You could use it to get map updates but not in real time.

[@dlvivlviv]

Tesla does not use its own self-driving system.

[@jamesc2388]

@@dlvivlviv yes it does. From hardware to software they make their tech in-house.

[@PhotoshopArt]

Video game is not driving me.

[@TheMegillahguerrilla]

Just because you can, doesn't necessarily mean you should. But technology has outpaced wisdom and is incapable of stopping itself

[@valorien1]

Just because you can (post clueless and idiotic comments), doesn't mean you should. Millions of people die in car accidents. If there's a technology we definitely SHOULD pursue, it's autonomous vehicles.

[@TheMegillahguerrilla]

@@valorien1 Thank you for losing your argument by making it into a personal attack. You seem to have a touching belief that technology cannot be corrupted by power. How romantically misleading is your faith in technology to conquer death.

[@valorien1]

@@TheMegillahguerrilla Oh, I had no idea you were the authority when it comes to who wins or loses an argument, specifically when arguing with you. You can't even tell the difference between mockery and a personal attack. I don't know you (nor do I wish to) enough to personally attack you, genius. As for your ignorant response: Any technology can (theoretically) be corrupted and made to harm people. That's something to be aware of, not a reason for all of us to go back living in caves. If you're so bothered, why don't you stop being a hypocrite and destroy your computer and smartphone? Also, please learn to read. I never said this will conquer death.

[@TheMegillahguerrilla]

@@valorien1 No, those aren't my rules, those are the rules of argumentation and debate. But never mind, as you were. Don't change

[@valorien1]

@@TheMegillahguerrilla Of course, I forgot there was a rule book issued by the U.N. in 1972. Give me a break, you made an idiotic argument and you got a suitable response. You don't win an argument just because people think you're a dumbass, but keep telling yourself that. I'll offer a piece of similar parting advice to you as well: please change. Be less stupid.

[@BenjaminGolding]

As someone with AI background, I call fraud on the instance segmentation at 11:00. Currently, there's no way to do instance segmentation in real-time that is faster than 4-5 frames a second. They processed the video after the fact and showed something (instance segmentation) that the car is not using at all. They are years behind Tesla from what I can tell.

[@TheKdcool]

Tesla FSD chip can do 2100 frames per second for 8 camera (260 each) I would would hope this demo car can process more than 5 fps

[@BenjaminGolding]

@@TheKdcool They are doing object classification, not instance segmentation.

[@jamesyoung8797]

@Jen are the Mobileye chips (hint, hint, Intel) still behind the old Tesla chips? BTW, Mobileye also uses 8 cameras for driving, the other 4 are for parking, which are crucial for FSD also, which Tesla does not have. Although I hope parking can be resolved with the help of the sensors, as shown by decade old examples. XPeng in China also has 3 extra cameras for parking only, which can be a concern for the Tesla's lack of parking cameras, but China's parking conditions can be far worse than those of Jerusalem.

[@jamesyoung8797]

@@BenjaminGolding Regardless though, this demo shows camera based FSD systems will work, bodes well for Tesla. Who is ahead of who is no longer my concern, it is just a matter of time. I think two of them should be partners again. FSD needs all the help we can get right now.

[@BenjaminGolding]

@@jamesyoung8797 If they are lying in the preview by showing instance segmentation instead of object classification which is what really occurring then who knows where they really are as partly it's fake.

[@ShanyGolan]

1. Tesla way better since it has more years of data-analysis of real road-driven cars, unlike intel's 1 car test. 2. Too many variables that mobileEye tries to take into account.. parking cars (why?), the popping and disappearing humans and color-coding them (white, red) will likely cause a disastrous car accident with them. 3. The driver must by law have his hands on the wheels at all-time. 4. The system takes WAY too much time analysis real traffic. No one in his right mind will wait 30secs to pass a simple intersection. 5. Intel should stop spending valuable money on failing projects and do what it "used to" know best: CPUs. (Thank you AMD anyway)

[@izybit]

@dimonat/videos Mobileye collects some data about the road, not the driving and no video.

[@eMKeaL]

You're being 100% clueless

[@izybit]

@@eMKeaL/videos That's you pal. Mobileye uses their REM tech to collect map data and nothing else. Feel free to prove me wrong though by posting a link that talks about other data they gather.

[@ShanyGolan]

eMKeaL yeah yeah. How many cars uses mobilEye? 0. Tesla? Yup, you got it.

[@eMKeaL]

@@ShanyGolan Nobody besides Tesla knows what kind of data they actually gather, what is the usefulness of the data and relevance towards FSD (if any). The point that they gather all those data for years and their cars have still same persistent issues in depth calculation and stationery object detection tells much. This parking feature is a big joke. They are nowhere near being close to full self driving.