Driverless Cars - Computerphile

[@user34274]

More of this guy please. Love his calm yet highly informative and precise delivery of information.

[@Праведныймиротворец]

You're an Indian what a coincidence

[@hil449]

@@Праведныймиротворец whats wrong with that?

[@Праведныймиротворец]

@@hil449 nothing wrong with supporting your countrymen but I just found that the opposite was the case

[@anonymic79]

This man needs to show up more often on Computerphile.

[@rafiahmed5017]

I really love computerphile ! You simplify things and make us understand using animations . love it !

[@SteveGouldinSpain]

I want to see autonomous cars that can drive in Cairo, Egypt, one of the few cities on the planet where even 'which side of the road you drive on' is an informed choice based on experience rather than civil regulations!

[@yousseffarid6830]

Most of 100k$ + cars here are sold without the most basic active safety equipment like even lane keeping assist, because it's utterly useless due to the non existing traffic regulations.

[@BeheadedKamikaze]

That sounds horrific. Why does nobody (a.k.a. the government) seek to educate the population?

[@yousseffarid6830]

@@BeheadedKamikaze they spend 4% of the country gdp on building new roads and bridges (one of the highest of any country on earth ), but they don't want to change traffic regulations. The maximum penalty for going 200mph is almost 15$ (so speed limits are almost symbolic), lanes are cosmetic only (they serve no purpose), dangerous overtaking, loud horns and exhaust modifications are totally legal + hitting and running isn't a crime. To put it in simpler terms, its totally uncivilised.

[@BeheadedKamikaze]

@@yousseffarid6830 I'm really sorry to hear that. I hope one day the people will implement policies to change that, and things will get better for everyone there.

[@ThisNameIsBanned]

The funniest part about self driving cars is, if they follow all the street rules and regulations, they get quite often deadlocked and have to stop and not drive at all, as everyone around them is not following the rules, so they cant rely on them (and to be safe, you have to stop or slow down to a crawl to guarantee you have enough room to break and avoid a crash). If you then program in to ignore some rules as you supposedly know better, it becomes an arms race again, and intentionally breaking rules because nobody really follows them.

[@ASilentS]

As you add more autonomous users to the system you'll eventually reach a tipping point where human drivers have to start driving with the robots and not against them just to get around.

[@vinny142]

@@ASilentS Chicken and egg. Electric cars have been around longer than the internal combustion engine but they did not take off until the problem of range was solved. Selfdriving cars will not take off until they can be truely autonomous and nolonger require a driver. That will *never* happen, simply because there are no companies that are willing to build a car and take full responsibility for every possible accident that can happen with it. And being an attendant in your own selfdriving car makes you pretty much useless because you will *always* be too late to notice that the computer is about to make a mistake. We know this, it happens all the time. The only place where selfdriving cars are of any use at all is on the highway where behavior is predicatble and humans fall asleep (and the fewet accidents happen because highway driving is easy) Selfdriving cars are very interesting from a machine-learning point of view but competely useless in every other respect.

[@MegaRad666]

I think you guys have forgotten the biggest problem in this industry, publicity. If a self-driving car is not following the rules (just like everyone else) and has a serious accident, the media will once again bring the hammer down on the industry as a whole. No human driver will follow the rules always, but enough --- uninformed --- people are brainwashed by the news media that a self driving car that ignored the laws one time and was unlucky enough to be in an accident is a problem. It is a marketing issue as much as it is a human behavior problem

[@AcornElectron]

Brake

[@brendonwood7595]

@@MegaRad666 It's a legal liability problem mostly. If the car has an accident the driver is responsible. Therefore the manufacturer will be responsible for all accidents there vehicles have. I suspect this is why all aircraft still have pilots in them. An autonomous aircraft is much easier to build since the rules are much simpler and better defined, and object avoidance is handled externally. They have been completely possible for 30 years or more but still we have pilots.

[@StrayCatInTheStreets]

A few weeks ago I finished my dissertation in deep reinforcement learning, imitation learning and offline learning in autonomous vehicles. If I'm being honest, driverless cars out in the real world in any condition without geofencing is many decades away

[@lucas29476]

Do you think so? I used to be interested in DRL too. Sent you a message on LinkedIn; would love to chat about this.

[@hil449]

thats a shame =/

[@black_platypus]

Can we get a video with Ramamoorthy and Moriarty together? I don't care what they talk about, it's just fun to say those names together ^^

[@HypnosisBear]

I love this guy. We need more videos with him. 👍

[@rjeffm1]

We talk a lot about what the technologists want/need. What do the insurers need to see before they allow a fully autonomous vehicle loose on the road?

[@TheFinagle]

All the insurance companies need is someone to point the law suites at so they dont have to pay out on claims. Once we clearly answer if its the owner or manufacturer thats always at fault for collisions caused by the AI they will be happy enough.

[@rjeffm1]

@@TheFinagle The problem is that with auto insurance, one insurer or the other always pays. With human driven cars, fault usually doesn't lie with the car, so individuals get tagged as bad risks and either can't get subsequent coverage, or it costs a lot more for that driver. With autonomous cars, the chain will always lead back to the builder of one or the other. In the limit, would anyone want to put their individual life in the hands of a machine designed to make decisions designed to minimize a given builders insurance liability?

[@Ziferten]

Quite enjoying these robotics and machine learning videos. As an EE this is much closer to my wheelhouse than some Computerphile content. More than anything I appreciate the effort put into making sure a variety of different topics are visited.

[@user-ue1vw6iv3s]

Thanks for commenting , I'll advise you look up to investing and making huge profit in Bitcoin with Samanthaleeward she's currently managing my crypto portfolio and making great return's*

[@user-ue1vw6iv3s]

On_Telegram @Samanthaleeward

[@pokegeaks2]

The largest hurdle for autonomous driving is coping with Canadian Winter

[@anonymic79]

I'm from Northern US states and "Where is the road?" can be a great family driving game certain times of year.

[@pokegeaks2]

@@anonymic79 And even when you find it the lines don't matter and everything is White or you find the road by identifying the section that is off-white lol

[@VineetSingh-rx9ni]

More such discussions can be welcoming

[@85set05]

Ah yes, the auto mouse 🐁 the fastest of all the land vehicle driving mice.

[@AudioPervert1]

The 6th mass extinction is going on ... And these folks want driverless cars... Makes sense zombie tech 🕳️

[@hil449]

@@AudioPervert1 yes we want because millions die every year on crashes

[@yttriumbagel]

It irks me that the car is driving on the left in the animations

[@davidgillies620]

I've been saying for years that the principal barrier to flying cars is not the physical platform but traffic control. You cannot expect people to move from a constrained 2D environment to an unconstrained 3D one. At a bare minimum, you would need real-time mesh-based communication between vehicles, probably with a central coordinator, and there would not be drivers (pilots?) in the usual sense. It would be a mix of almost complete autonomy from direct human control at the system level mixed with near total lack of autonomy at the vehicle level.

[@AnotherPointOfView944]

Also, single point of failure. Not addressed here. You need multiple sources and multiple receivers and processing units to make this safe.

[@andrasbiro3007]

Already been done. Tesla uses redundancy from the cameras to the processing units.

[@ceptimus]

Any human driver is also a single point of failure. The autonomous systems just need to demonstrate that, on average, they are safer than human drivers: when that happens, and the technology becomes cheap enough, then autonomous vehicles will rapidly become commonplace. It's likely that eventually, human drivers will be banned from driving on most public roads, because they represent increased danger compared to the self-driving cars that, by then, dominate.

[@udaynarayanmukherjee5208]

Am I the only one seeing the piano sheet on the wall ... What peice is that

[@ceruchi2084]

What a great guest!

[@mattj65816]

A jacked-up pickup truck is likely to do something obnoxious, a Prius is likely to do something stupid, a minivan is equally likely to do anything at any given time--it's completely random. That kind of thing literally *is* in my mind as I'm driving. How do you program for it?

[@Megaranator]

you drive defensively also ai can just learn these "stereotypes" better than any human and adjust acordingly

[@mattj65816]

@@Megaranator Yes, but I didn't want to do the "AI will learn it" punt--if Computerphile had done that, this video would have been five seconds long.

[@euchale]

@@Megaranator How can I opt out to be on the same street as a car who hasn´t learnt these stereotypes yet?

[@andrasbiro3007]

​@@euchale You can't. There are already like 1.5 million self-driving cars on the roads. Of course currently they are only allowed to drive under strict human supervision.

[@MinedMaker]

Theres a video out there of a self-driving Tesla being very confused and thinking it's passing millions of stop signs on the highway because some road workers were driving in front of them with signs in the back. Proves you will never be able to predict all the edge cases.

[@anonymic79]

That's funny, kind of like Toyota drivers.

[@Thepewdiepiebro5]

Is this autopilot (they're lvl 2 system for highways) or their bets program "FSD beta", which has much more capabilities. Also, those edge cases are just because of the lack of data.

[@pequalsnpsquared2852]

Meanwhiile humans, who in theory should be able to deal with this, still drive tired, drunk, and distracted, and can't see in all directions at once. On balance I think I can tolerate an edge case once in a while

[@m_sedziwoj]

It would happen once never again, and people making same mistakes many times. Even if you eliminate unintended acceleration (driver press acceleration, when he want brake), is enough to compensate this one situation multiple times. It is not when AI makes mistakes, and humans don't, it is about who make less mistakes in long run.

[@kristoffseisler2163]

I miss Mike Pound its been a while since i last saw the hero

[@samuelthecamel]

Coudn't you just feed the computer a bunch of videos of people driving and train it on that?

[@funnymatt]

They sort of have- that's what Tesla and Comma AI have done.

[@lukas.bekermann]

@@funnymatt are doing*

[@boiledham]

16:45 SLAM! C: Great video and an awesome introduction to the state of the field of self-driving cars in the real world. Anyone interested in pursuing a career in autonomous transportation can start their education now and be ready to enter a massive industry. This problem among others requires engineers and artists to work together in order to cultivate a planet where technology, nature, and human life are not seen as separate factions bullying for control of finite resources.

[@alexpetrovich85]

We're not getting fully autonomous AI cars. At best, you'll get autonomous cruise control and self-parking, but the amount of data processing, storage, and infrastructure required for the majority of cars to operate fully autonomously at human levels (let alone superhuman) is magnitudes beyond anything achievable in the near future. This hype will lead to another AI Winter. Additionally, vehicular policies will be implemented to dumb down, limit, and simplify travel so that it can match the compatibility and compacity of these AI models.

[@GroundThing]

Agreed, though I feel like this is going to be one of those "lessons learned in blood". There have already been instances of people killed by autonomous vehicles, but in order to get to the place where people, or at least regulators, understand that full autonomy is impossible, at least for the foreseeable future, we're going to see a lot more deaths and serious injuries before we get to where autonomy gets limited to cruise control and parking.

[@hil449]

im hoping you're wrong, way too many people die on crashes today, humans are not meant to drive

[@alexpetrovich85]

@@hil449 True. But then we'll have to deal with the few but tragic automated deaths which will statistically occur with the self-driving cars. The suffering of accidents will always exist, and there is the question of what price we are willing to sacrifice to mitigate them (freewill, mobility, regulations, cultural shifts etc.). Personally I hate driving, so I wouldn't mind having a auto-chauffeur safely drive me to the store and back; just weary of the hype around what is being promised to the public.

[@hil449]

​@@alexpetrovich85 of course there'll be crashes, but im pretty sure it'll be much less. Maybe just a fraction of what we have today. Im not saying everyone should stop driving (that would be wishful thinking at least in this century). The sad thing is the few automated deaths will be used by the media to attack the technology even if in the long run it saves thousands of lives. Humans are so irrational sometimes, i think its the biggest barrier for this technology =/

[@recklessroges]

So "10 years away" still?

[@dfmayes]

Ha ha. Just like fusion power.

[@MondayHopscotch]

It's not just student drivers that are terrible at roundabouts... Here in the US, people are amazingly bad at roundabouts. Stopping before entering an empty roundabout, making left turns from the outside lane, stopping in the roundabout to let other cars in. It's maddening.

[@prithviquantum3777]

Awesome

[@AcornElectron]

6:12 exactly that Brady.

[@kuhluhOG]

Here a more ethical question: Who is going to take responsibility when a self-driving car actually drives somebody to death?

[@brendonwood7595]

That's a legal question, and the answer is the manufacturer. This is why I don't think they will happen anytime soon. It's not just a technical problem.

[@kuhluhOG]

@@brendonwood7595 well, imo laws should represent the morals and culture of its people so yes, it definitely is a moral question; especially if it's a topic like that

[@KU-mg9el]

If statistically beneficial who cares

[@kuhluhOG]

@@KU-mg9el quite a lot of people: If it's the car manufacturer, a lot of them are going to be hesitant about making the cars fully autonomous because even if it's rare, they have A LOT of customers. If it's the owner, well, I know a lot of people who would then drive themselves instead (and/or not own such a car). If it's the individual programmer, they are going to struggle to fill the positions (hey, programmers are still far fewer than available job positions). You can't really do nobody either in a lot of areas (cultural thing).

[@brendonwood7595]

@@KU-mg9el The companies that will go broke from the liability. Currently the aircraft manufacturers, as a self flying plane is already a technical reality but the pilots are still there. As long as the legal system assigns blame and payouts then self driving cars aren't going to happen because no company can afford to sell them without them being completely perfect, and that is not possible. Hoe long do you think a car manufacturer would last if they had to pay to repair every car in an accident? Even if only 10% of the current accident rate? And if anyone dies then that is a huge payout.

[@sriadityas1042]

Amazing Content Guys...

[@yoyoyogames9527]

really good and interesting video

[@BeheadedKamikaze]

Nice bass on the wall Sean! Play it for us? :)

[@kyrianrahimatulla1561]

Going through a comment section at a moment when not enough time has past for comments to get a lot of likes or dislikes and thus to not be sorted by popularity is a weird experience.

[@googleuser7771]

Yeah you're not being told what to agree with implicitly by how many likes comments have. Instead you have to read each comment and make your own mind up about whether you agree or not

[@zizkazenit7885]

I’m excited that autonomous vehicles are just around the corner! And have been for the last ten years

[@Richardincancale]

Arghh - someone needs to tell the robot which side of the road to drive on!

[@derstoffausdemderjoghurtis]

*confused car noises*

[@kennethkath6527]

I believe automated cars are/will be programmed to avoid pot holes. So, assuming there's a pothole in the middle of the intersection and assuming there's multiple automated cars, will these automated cars try to avoid the pot hole which may lead to accidents or will it go through the pot hole and save lives?

[@andrasbiro3007]

Not programmed, that's the point. They have to learn from examples, because the required software would be many orders of magnitude more complex than what human engineers can write.

[@AcornElectron]

If EVERYONE has them then the system becomes exponentially more reliable immediately.

[@DumbledoreMcCracken]

That is a utopian vision. Hackers aren't a problem when people drive (mostly)

[@VincentGroenewold]

@@DumbledoreMcCracken Why would that be utopian. It's possible, not even that hard. Companies just need to have the technology being far enough that it can be added to all their new cars (just like cars had other gadgets in the past), then after they all gather their data, they can start rolling it out et presto. After that, when the degree of implementation is high enough, they can cooperate and use a standard to make cars communicate with each other. I don't see how that is not the future to be honest.

[@DumbledoreMcCracken]

@@VincentGroenewold Not easy at all. You never have written any guidance software. I had thought that having the cars communicate would solve most of the problems, but it doesn't. The animal targets and pedestrians, together with man-made accidentally created debris, makes the whole thing collapse in chaos in other than pristine laboratory conditions. Never going to work.

[@Furiends]

​@@VincentGroenewold The way its phrased at 15:36 seems to suggest they are talking about coordination but thats not actually what Ram responds with. He instead mentions adding instruments to sighs and roads. I can also fill in that he's say cars as well except that thats an already solved problem in computer vision. Ironically signs and the road are not. What he doesn't mention in this discussion is that the value of adding instruments to signs and roads is that they imply something about the conformance of the road. It's not an unmaintained road or a data point on a map that may not actually exist in reality but is a conformance proof to the AI. Ram states "its only a matter of time before something like this *also* comes in" he's not saying its a replacement or a panacea just that it would help to solve a bunch of problems with self driving. Notice he's talking about the this within the context of rebuilding infrastructure or in other words this goes along with making "self driving ready" roads

[@hil449]

@@DumbledoreMcCracken never say never

[@jmalmsten]

For me the real tipping point will be when manufacturers take on the responsibilities of drivers. When you can buy a car and legally use it without knowing how to drive a car or even be awake during the drive. When it is so safe that you basically have to sign a user agreement when you as a human take the wheel and have to confirm you take back the responsibility for any accident. That's when it has become safe enough. Because these manufacturers will be soooooo hesitant to be liable to accidents, injuries ir even deaths. They will be so paranoid to take that on. Only then will I fully trust the autonomous driver. I believe it will happen. And I am optimistict about the prospect. I want it to happen. Yes, sure. A mechanical emergency brake lever should always be available for when stuff go awry. But. I do think it will come to a point where having a human driver is looked at as some kind of Evel Knievel stunt shenanigan by fellow road people.

[@user-ue1vw6iv3s]

Thanks for commenting , I'll advise you look up to investing and making huge profit in Bitcoin with Samanthaleeward she's currently managing my crypto portfolio and making great return's*

[@jmalmsten]

Ok. Now I am confused. Has Computerphile been hacked or am I just being trolled? :P

[@tomvleeuwen]

Why are all the animations mirrored? Something British?

[@mountp1391]

Thank you.

[@user-ue1vw6iv3s]

Thanks for commenting , I'll advise you look up to investing and making huge profit in Bitcoin with Samanthaleeward she's currently managing my crypto portfolio and making great return's*

[@yrks1109]

His Voice is close to Sundar Pichai's voice!

[@codycast]

Sounds like the dude who helped me troubleshoot my router

[@hil449]

@@codycast sounds like you're r4cist. Reported

[@jti107]

interesting idea to use game theory to model the pedestrians

[@alishmanandhar3444]

I got here at 39 sec and there is already someone who disliked the video what kind of hater is that

[@XavierAway]

jealous haters

[@piyush123ization]

Haha, true!

[@sergio3674]

@Gnome Whyte why? If computers can drive better and save lives I want them driving.

[@kennethkath6527]

@@sergio3674 true but huge amount of reasonings and time needs to be added for the car to decide what decision to make. There's still many loopholes in driveless cars but I believe things will progress

[@igorthelight]

@@sergio3674 Agree

[@JuicyJonesHQ]

I'm thinking many drivers couldn't figure out where they are on a map. 🤣🤣

[@treyquattro]

unless you entirely standardize and control all aspects of the road traffic network, it strikes me that self-driving - and interacting, with humans amongst other considerations - requires AGI, which probably won't be here until the end of this century

[@tjejojyj]

Self driving vehicles already communicate with each other through brake lights, turning signals and their camera systems, just like humans. Won’t FSD-logic be incorporated in cars as safety measures to stop drivers doing something dangerous, like the way electronic stability control can “take over”? The question I would like to hear posed is whether there is a chance of the full self driving project failing or being limited to particular domains (geographic and driving conditions). Up to 2018 about $80 billion had been spent and the over confident announcements are its introduction are getting pushed further back. What is the acceptable level of edge cases? If self driving cars are safer than 95% of human drivers, isn’t it better that some people don’t drive, except where the AI fails? Finally, wouldn’t the field advance faster if the different developers shared their edge cases with each other?

[@Eevee13-xo]

Automated cars would make police chases a lot easier

[@kennethkath6527]

assuming all cars on the freeway is automated.

[@Eevee13-xo]

@@kennethkath6527 that's exactly what I'm assuming brother

[@jamesmorgan3623]

@@kennethkath6527 I think they will be; insurance costs will make "manual" cars unaffordable in the end.

[@EgoShredder]

No need for Police in the dystopian future coming our way in the next 20 years or so. Armed flying drones connected to the internet network, can go straight to the scene and engage or take out the person being non-compliant to the societal system of control. Before this a persons access to essential services and their own money / food, can be withdrawn, until full compliance is observed again. Technology is great.................... yes?

[@googleuser7771]

There won't be enough freedom to even physically be able to break the law in the first place

[@dfmayes]

I worked in the software industry for 35 years so I know how well engineers write code. No way I'm trusting a "software car."

[@hil449]

its not software development. Its data science, totally different thing

[@dfmayes]

@@hil449 those cars have millions of lines of code.

[@tohamy1194]

Interesting

[@AkanoWire]

istn there a gpu build for autonomous driving? nvidia drive or smtn like that? xD

[@TheFinagle]

Country level variances can be easily solved by having it available locally. The same system that will be telling the AI cars when lights are red or what the speed limits are could also deliver (as a initializer packet) a set of behavioral expectations for the roads. There will be a lot of infrastructure updates put in place BEFORE AI cars can be deployed. The starting point can be to add these things for the fancy dynamic displays in new cars - think having the speed limit listed on your dash next to the speed

[@DumbledoreMcCracken]

Self driving is an impossible goal in all but a theoretical utopia.

[@BeheadedKamikaze]

I guess you never heard of companies like Tesla, Mazda, Waymo, AMD, Intel, nVidia, Boston Dynamics, etc. etc.

[@DumbledoreMcCracken]

@@BeheadedKamikaze which have all failed to deliver. You have proved my point exactly: that only the fan-bois believe the malarky.

[@BeheadedKamikaze]

@@DumbledoreMcCracken "Failed to deliver"? What on earth are you talking about? They are making great progress everyday. The work is currently ongoing, and you have not presented any reason to back up your assertion that it can never be completed. You also never made such a point about anybody believing anything. What you actually said was, "Self driving it[sic] an impossible goal in all but a theoretical utopia." - and simply claiming that something is true, does not make it so.

[@user-221i]

Exactly. Cities should invest in public transport.

[@a-very-monday-tuesday]

computers are cool

[@alokkumaryadav6209]

Who is from india

[@fraac]

a city like bangkok should ban all human drivers and mandate that the robot cars be networked. instantly the safest streets in the world with the best public transport

[@Patapom3]

You don't don *any* assumption that would assume citizens will obey the law. Period!

[@Poxenium]

Really? Nobody mentions the industry leader? 🤔 Look up Andrej Karpathy's lectures 👍

[@AshleyWilsonAU]

If you meant Tesla, Ram did mention their success with making custom chips / hardware.

[@user-221i]

Lol. Tesla is a scam. They have accidents constantly.

[@markboz3366]

So glad I'm at an age that I probably won't see this. I can imagine the first few years being full of court cases.

[@xyZenTV]

You want to die just to miss out on coverage of court cases? Damn...

[@hil449]

im so glad im at an age i will see this. I cant happen to see less people dying in crashes worldwide

[@aqueousone]

10:00 Suggesting that anyone who has seriously driven (whatever that means) would agree that the “act of driving is fundamentally an act of conflict” shows a complete misunderstanding of what mindset is required to safely and consistently drive over 100,000 miles per year.

[@Richardincancale]

Try driving in Paris!

[@hakanolausson1367]

I'd say that IS the mindset of driving safely. Me trying to cross a road at the same time as a car is trying to pass that part of the road means that it is a conflict of interests if ones does not realize that there will be conflicting interests and takes that into consideration then there will be a less safe environment. A different example would be: Deer: "I'm going to run to that green pasture over there" - Driver: "I'm going to go pick up my kids" and it happens that they would want to pass the same point of the road at the same time but in different directions. Which are conflicting interests (if one wants it to be safe for all parties) and if self driving cars cannot safely handle conflicts then they will be unsafe or extremely slow and the ability to compromise will be vital.

[@Megaranator]

way to miss the whole point :D

[@Zeecarver]

So, the ultimate goal is to remove all human drivers, is it?

[@xyZenTV]

Ideally, yes, human drivers are awful. Realistically, that will never happen.

[@funnymatt]

I have damn good Level 2 autonomous function with my Comma 2, which is essentially a cel phone stuck to my windshield. Self driving cars will be solved soon.

[@alenpaul2523]

Love from India .

[@kyleblazer1912]

if (GoingToCrash) { dont; }

[@ThisNameIsBanned]

This is way too broad and explains so little. Better with some examples and how they are solved, instead of just saying "The big players have solutions" and wave the question away with that. Interesting topic regardless, as its a pretty large field of very different technologies playing together to solve many issues in many different ways (be it by optics alone, or with data from infrared cameras, or like a radar and so on).

[@mathewsjoy8464]

Go do a degree in it lmao they don’t have time or any need to go into detail

[@ZandarKoad]

Agreed. The big players need to appear as though they have solutions to most internal and external observers for continued funding. What they actually have is seldom open to full, independent investigation.

[@seasong7655]

Maybe this is another problem, where people only think of additive solutions. We could develop a system for several billions and put 50 years of research into it, and then add it to our cars, OR alternatively we could just not buy cars, which would be the much easier and cheaper subtractive solution.

[@yashvashistha6004]

What would replace cars?

[@ciherrera]

Have you watched the recent ASAP science video?

[@Diggnuts]

It's like these people live in some reality where FSD beta does not exist!

[@m_sedziwoj]

That you not mention Tesla until 3/4 of video, this is showing you are bias (to not mention anyone, but this way you ignoring reality), sorry, but all about gathering edge cases is what is Tesla doing, and they are nice example how it can be done in real world. 5:30 nope, someone not fallow what Andrej Karpathy is doing... 10:15 it is not act of conflict, maybe in India, but in normal countries as in EU or USA driver want to get to his destination, not to confront others, and law and logic showing that organized movement is fasted (in limited space and knowledge situation) 13:00 yeah, if you use Nvidia hardware which use 500W is pointless, but this is why is only should be use in development phase, and switch to dedicated when scope is known.

[@TheManAshley]

He doesn’t broach the issue of the elimination of human drivers and going full driverless. The simplicity of line following cars will eliminate the need for overly complex computer systems. Actually a bit more complex, but not much. A full driverless society would see the reduction in vehicle numbers by 95% or more with each being part of a fleet that would make 50 or more runs a day at a cost of $.50 (in current dollars) each or half or less with shared occupancy. Waiting time could be a couple of minutes. Traffic deaths in the U.S. in 2020 we’re around 42,000. With a full driverless system this total could be 420. There is no system with a vested interest to save 42,000 lives a year, eliminate all driver jobs, save $450 billion in annual highway infrastructure building and maintenance, or deal with a 95% reduction in all jobs. A raspberry pi could handle the navigation of driverless cars. I’m still working on my revolutionary theory that the earth actually revolves around the sun, so when I get the math down I’ll put out a paper.

[@chinofiron]

Oh wow first 🥇

[@rohankokatanur6426]

Only i understood this video conpletely without subtitles . Other than Indian accents are hard to get , plz try to add subtitles..

[@vorrnth8734]

Please don't call autonomous cars self driving. Simply because cars are automobiles which means self driving. Self driving cars would be self driving self drivers.

[@hakanolausson1367]

Auto mobile means self moving (-ish), not self driving and a that something is moving by itself does not mean that it is in control of it's movement. Autonomous driving would mean being in control too.

[@Gringohuevon]

boring

[@truckcompany]

This guy was annoying, he didn't answer some of the questions. He didn't answer the question at 5:17.

[@cheddarfodder]

I'm not sure this guy realises how much progress Tesla have actually made on this. They already have 1mn+ vehicles running FSD in shadow mode which send all the data back to their super computer (number 5 in the world by flops). Nobody is going to catch them. He lost me when he mentioned radar and lidar, they aren't needed. Cool talk though.

[@valdiste123]

Seems like Tesla marketing has done it’s job on you

[@cheddarfodder]

@@valdiste123 let's come back in a couple of years pal. We'll see who's right, I've done my research.

[@AshleyWilsonAU]

@@cheddarfodder By research, you mean following Elon on Twitter and watching event streams? Lol. Tesla has indeed made a lot of progress, but they're still years away from having a system that is as versatile as a human, not just safer statistically. FSD has been "coming soon" since 2016.

[@cheddarfodder]

@@AshleyWilsonAU by research I mean hundreds of hours of content: books, interviews, podcasts, 3rd party research, events. It's something that I'm really interested in. You might know this stuff but I'm going to mention it anyway just in case... Remember how Google became the dominant search engine by allowing every user interaction to improve the system? That's precisely what the guys at Tesla are doing with their data to train their neutral net. They already have billions of miles of real world data. Their software is running in the background on most vehicles and every time a human action conflicts with Tesla's FSD, it's logged and they study it. To be fair, I'm not sure we'll see autonomous in downtown Mumbai any time soon... but in many other countries, we'll see robotaxi's way faster than you think.

[@taragnor]

@@cheddarfodder : Judging by Elon Musk's history with backing up his exaggerated claims, I'd bet on the other guy being right and not you. Research Hyperloop, the Las Vegas Loop, Solar Roofs, battery powered semi-trucks and Neuralink. Bunch of outlandish claims made with zero backing up. The Las Vegas loop ended up being a Tesla in a tunnel going 30 mph, instead of the ultra-high-speed "wormhole" it was originally sold as. Oh and the Teslas in the tunnel aren't even autonomously driven, they need drivers.

[@AudioPervert1]

The 6th mass extinction is going on ... And these folks want driverless cars... Makes sense zombie tech 🕳️