Forums

Is full self driving really necessary?

Is full self driving really necessary?

I don’t really think full autonomous driving will be possible anytime soon and actually doubt the necessity of it anyway. AP has a few quirks that would be easy to fix now and then id say it’s done and call it co-pilot and id be happy.

carlk | 12 juillet 2019

Like most things in life it's nice to have but not necessary.

jjgunn | 13 juillet 2019

Horse hockey....

It's definitely necessary because less accidents & deaths on the road.

andy | 13 juillet 2019

I suspect that, in a few years, it will become difficult to get insurance without it.

Xerogas | 13 juillet 2019

There will come a time when manually driving a car will become quaint, and after that it will become illegal.

spuzzz123 | 13 juillet 2019

I think I’d be happy if they could give me freeway autonomous driving. No nags and no need to pay attention until freeway route ends. Feels like that should be attainable. City driving so much more complicated.

mcmack15 | 13 juillet 2019

I doubt I will see it in my life time, which is fine with me. I do think lawsuits from deadly accidents during the 'learning curve period' will delay things for a few years as well.

To follow up on Andy's post above-----at some point I believe it will be mandatory for all the data captured to be available to the insurance companies, and then the fun of driving will be forever over. They will know when you didn't use your blinker; when you didn't come to a complete, 100% stop at a stop sign; how many times you exceeded the speed limit by 1 mph, etc,. etc. and they will charge you accordingly.

Madatgascar | 13 juillet 2019

The driver assistance systems are already very reliable and will only get better. It has been shown that humans are inherently very bad at monitoring highly reliable systems. That’s why FSD has to make it all the way to L5 autonomy at a level of safety higher than that of a sober, competent human driver (taking drunk, distracted, enraged, sleepy, and otherwise debilitated humans out of the equation). It’s a very high standard, Elon’s 99.9999 is probably right.

They may get there, but my guess is they will need to err on the side of safety so heavily, it will really slow things down. I probably won’t have the patience to be driven so “safely.”

mcmack15 | 13 juillet 2019

Fish, amusing you once again throw a dig at someone responding---I do wonder if you could go two weeks with only positive posts-------no insults; no digs at folks; no hijacking posts and complaining about AP for the millionth time; etc.---only positive posts. If you can do it I will buy you a cold beer if we ever meet up..

In any event, unlike you, I always come to complete stops at the red lights and stop signs---so we both would see our insurance rates sky rocket. I would get hit with insurance increases because of the highway driving. Most of my mileage is from highway driving around the greater Boston, MA area where the speed limit is 55 mph. I am guessing and I believe anyone on this Forum from MA will agree) that less than 1% of the folks driving on routes 128, 93, or 95 drive at 55 in the right hand slow lane. And nobody in the other 2-3 lanes ever goes 55. If you did, you would cause such traffic backups and so many accidents as the other 99+% tried to get around you (and they would, and they would also purposely cut you off as they got back into the lane in front of you). Heck, the police don't even drive 55 on these roads most of the time.

Bottom line, my insurance would be through the roof just for driving in the flow of traffic.

And yes, I do have fun with my Tesla, which includes using the excellent acceleration to pull onto the highway, or to switch lanes on the highway. It is one of the main reasons I bought a Tesla---for the performance.

Madatgascar | 13 juillet 2019

@Fish, I don’t get the sudden stopping problem, I just find that NOA is too polite to compete with aggro LA drivers, even on Mad Max mode.

I agree that working L5 autonomy is more likely to eventually be an accomplishment of the collective (multiple manufacturers working with improvements in the infrastructure) than a triumph of the individual (Musk/Tesla).

For now, it seems to me Tesla is in the lead. I’m watching the other systems, but don’t think they will have the chance to be as good. AI learning can really have big breakthroughs. The rest of the industry will benefit from Tesla’s pioneering, and eventually we will have a system that works. Just hope Tesla doesn’t take too many arrows in the process.

calvin940 | 13 juillet 2019

I think FSD is valuable and will become necessary. As such, companies need to start working on it now to get it ready for when it is necessary.

andy | 13 juillet 2019

The real breakthrough will come when vehicles communicate with each other and that is also the point when legacy cars will find it difficult to mix.

There are plenty of models in the communications world that could be used as examples - either from OSI or RFCs.

Cars are like data packets and roads like Ethernet cables. CSMA/CD seems, to me, to fit perfectly as an example - the difference being that the data packet has the intelligence rather than the transmitters and receivers.

Make it mandatory for all road vehicles to carry some form of IoT tag and a lot of the complexity would go away - and I it would also accelerate the point as which legacy cars are cleared from the roads as the best you could do with them would be to retrofit some sort of accelerator and brake override.

WantMY | 13 juillet 2019

It is lucrative for taxi cabs, trackers and etc where driver is an expense. As for consumers, it is not important as we could drive and some may even have a pleasure driving their car, lol

mcmack15 | 13 juillet 2019

Andy +1000%

Great input-------I think you are 100% on the money.

kaffine | 13 juillet 2019

andy

Cars are like data packets and roads like Ethernet cables. CSMA/CD seems, to me, to fit perfectly as an example - the difference being that the data packet has the intelligence rather than the transmitters and receivers.

With cars I would prefer CSMA/CA. I think Collision Avoidance is preferred over Collision Detection in the case of cars.

vmulla | 13 juillet 2019

@kaffine,
Tell me you're not a network engineer ;)

Is self-driving really necessary? I don't know about necessary, but most of Tesla's popularity is because of its advances in this area - if that's a measure yes it's very much wanted.

Hp.1193 | 13 juillet 2019

Self-driving with Boring Tunnels throughout major cities in the US would be gold.

lbowroom | 13 juillet 2019

Some more nonsense in here. You can use TACC without autosteer but you can't engage autosteer manually without TACC.

ReD eXiLe ms us | 13 juillet 2019

Hal Fisher opened with, "I don’t really think full autonomous driving will be possible anytime soon..."

Luckily, it doesn't MATTER what YOU 'DON'T THINK'. The NHTSA has spoken on this matter. Autonomous driving WILL happen. For precisely the reason intimated by jjgunn.

NHTSA | Driver Assistance Technologies
https://www.nhtsa.gov/equipment/safety-technologies

"In 2017, 37,133 people died in motor vehicle crashes. Research shows that the vast number of vehicle crashes are tied to human error. New driver assistance technologies hold the potential to reduce the number of crashes and save thousands of lives a year. We here at NHTSA are committed to making our roads safer for everyone."

That 'human error' ties to the factors mentioned by Madatgascar. Autonomous cars will not get tired, sleepy, exhausted, fatigued, distracted, emotionally upset, high, drunk, or otherwise intoxicated, and to my mind, autonomous vehicles would not choose to operate during inclimate weather conditions either. All situations typically brought on by faulty decision making by human beings, who do all of the above as an expression of 'freedom' and end up dead or killing or injuring someone else as a result. Sure, you can be 'free' and you can 'have fun' but you must also suffer the consequences of your actions and decisions.

Please, do offer alternate viewpoints, explaining why it is a good idea that people be allowed to drive when they are sleepy, intoxicated, under the influence of drugs, emotionally distressed, or during inclimate weather conditions, because they 'know better than a computer' and deserve to exercise their 'freedoms' for the sake of 'fun'. So many people endorse the 'Live Fast, Die Young' philosophy, yet forget all about the 'dying young' part. Then they want to act surprised to find they are slowly wasting away due to Cancer, Emphysema, and old age.

RES IPSA | 13 juillet 2019

Until all (or at least a vast majority) are Level 4-5 autonomous, it will be annoyingly slow and overly cautious to occupy a fully autonomous vehicle. It would only make sense if you have plenty of time to get to where you are going or have been consuming alcohol and/or sleepy. It would also be nice on long road trips.

When all (or at least a vast majority) are autonomous, car will be communicating with one another, driving very fast with perfect timing... it will be awesome to seamlessly merge and change lanes at 120 mph.

Magic 8 Ball | 13 juillet 2019

Parity or better:

"Nearly 1.25 million people die in road crashes each year, on average 3,287 deaths a day. An additional 20-50 million are injured or disabled. More than half of all road traffic deaths occur among young adults ages 15-44."

Hp.1193 | 13 juillet 2019

Cars communicating with each other, in addition to FSD, will be huge.

RES IPSA | 13 juillet 2019

I am not interested in Level 4-5 autonomous driving in urban areas until the system can drive more efficiently than I can drive on my own.

Long road trips is a different story, but Level 3 can handle that.

RES IPSA | 13 juillet 2019

When autonomous cars are wide spread, people will start buying autonomous aerial drones for short range personal travel... the future of travel is in the air... not on terra firma

Hp.1193 | 13 juillet 2019

Waze, google, Apple have cars communicating or phones ?? Hmmm

Omicron | 13 juillet 2019

With each new safety and convenience innovation, you can ask is it necessary. Seatbelts haven't always been available, and now it's hard to imagine not using them.

Personally, I can't wait for FSD, but that's not a "Necessity."

I do know several people that can't drive due to various issues, and I am sure everyone knows someone that shouldn't be driving anymore, but they just won't quit.

FSD would be a game changer for these people.

thedrisin | 13 juillet 2019

There is already FSD. It is called Uber and Lyft.

Lonestar10_1999 | 13 juillet 2019

Ford and Volkswagen just inked a multi-million dollar deal to co-develop autonomous driving technology.

I realize that there are intellectual property concerns with this holy grail of artificial intelligence but I think for best chance of commercial success, full collaboration is required. It would be great if developers of FSD, regardless of their employer, could share their research. This is the approach medical research takes with goals such as curing cancer, AIDS, and other enormous challenges.

kaffine | 13 juillet 2019

vmulla : No I'm not a network engineer. I'm an electronics tech and try and stay away from computers and networking.

vmulla | 13 juillet 2019

@kaffine,
Your analogy made me guess :)

kaffine | 13 juillet 2019

Lonestar10_1999:

The issue with collaborations is how long they take to do anything.

The old saying

If you need something done now do it yourself, if you have time delegate if you have forever form a committee.

Look what happened with EV charger standards.

thedrisin | 13 juillet 2019

@Lonestar10. FSD is a proprietary Tesla acronym. The actual equivalent of automation is not actually defined. However, high level automaion, level 4 or 5, will probably need a concerted effort.

vmulla | 13 juillet 2019

@thedrisin,
What? FSD is a Tesla proprietary acronym?
Didn't think from that angle ever.

Earl and Nagin ... | 15 juillet 2019

@Andy and kaffine,
I respectfully suggest that CSMA/CD and ethernet protocols are totally the wrong thing for V2V communications. They are non-deterministic, slow to respond, subject to total collapse from statistically likely confluence of events (such as a bad thing on a highway where all cars would need to communicate at the exact same time) and reliable CD (collision detect) is nearly impossible for wireless systems, unlike wired (like ethernet) systems.
Instead, I believe that some sort of synchronized scheme would be necessary to ensure critical radio communications can happen in a reasonable amount of time.
Personally, I have more faith in free-space optics instead of radio or RF for V2V, with position sense multiple access as we've been doing for over a century. In other words, lets use light signals and colors to indicate intentions such as turn signals, brake lights, backup lights, etc. These are simple things that a camera can easily see for all vehicles that could affect each other. Adding a bit of pulsed coding to carry digital info might be helpful to augment the spatial/wavelength/coding (aka position/color/blinking) that we currently use. multiple vehicles can use the same signals at the same time and a simple, cheap camera can separate the vehicles by their relative angular position, thus tracking every nearby vehicle, even those being manually driven.

carlk | 15 juillet 2019

@Earl and Nagin

Exactly the same thinking here. An autonomous system that relies on V2V will never work. It will require ALL vehicles on the road to have that ability which is an impossible thing to achieve. Even if governments mandate that on every new car, which is also unlike to happen, it will still take decades to replace all older cars that are already there. Just one car at the intersection that could not communicate with you will make the system to fail. And if you can take care of that one car your system already not needing that to do that job. Your point that it will need a fail proof real time communication of course would be another show stopper. It might help in pseudo autonomous case when only one type cars running in a small community but never for the general case.

The system has to work in a way just like how you and I are driving our cars. Be able to recognize road signs, traffic lights etc. is the basic minimum. The toughest challenge is how to recognize other cars' intention like at a 4 way stop or when they are trying to change lanes without turning on the flicker. To a certain degree we all know how to handle these situations. In the 4/22 presentation Karpathy mentioned this is one thing Tesla is doing for its system to do too. He said Tesla cars can observe small movements (body languages) of car and pedestrians to predict their intentions such as wanting to change lanes or to move into the crosswalk. They will then observe their actual move to compare to what were predicted and use the result to improve the NN. All these are done in the shadow mode whether the AP is engaged or not. I don't think any other company can do this since none of them has the deep learning implementation. This is probably the final piece of puzzle and ability to solve that will separate Tesla from everyone else.

walnotr | 15 juillet 2019

E&N, a brilliant idea! Having a standard defining pulsed lights on vehicles transmitting data about what they are doing would make using cameras to get more than visual clues even more useful. Naturally, the cameras would require a fast enough frame rate, but significant data could be transmitted even at a 60Hz frame rate. Speed delta could be transmitted and used to determine the braking force needed thereby smoothing out deceleration and acceleration. Likewise, traffic lights could do a similar thing to announce upcoming changes and/or types of intersection being approached.

I’m sure the list could be expanded to include other edge cases.

andy | 15 juillet 2019

I like the way this thread has developed - thanks for the ideas, they make good reading.

Techy James | 15 juillet 2019

Due we need FSD or autonomous Level 5 cars. My answer is simply Yes.
Lets look at the facts in 2018 there was 36750 deaths from car crashes in the US. In addition 4.5 Million people was seriously injured. 1.6 Million crashes in 2018 was cited Texting and Driving at contributing factor. That is 25% of the reports accidents. From just one form of distracted driving. Then you add DUI which is listed as causing 31% of accidents reported. Just those two types of bad human behavior accounts for 56% of all accidents.
We introduce FSD and lets say it's just 99.9% error proof. That still means 36,713 more people are alive each year from traffic deaths. and 4,495,500 don't get injured each year. Yes we still have 37 people die each year from Traffic accidents and there is still more room for improvement. But think with FSD and inter car communication, people will get to their destination faster than they would have without FSD due to few accidents to clog up rush hour, and cars communicating together to get each passenger from point to point faster than ever before.
Yes all this is great, but there is some downsides to this. Loss of jobs as cab drivers and rideshare drivers are now replaced by FSD. Fewer needed law enforcement as FSD driving cars would be programmed to follow local laws, and could even identify erratic behavior from Self Driven or faulty FSD programming (i.e. hacked either by owner or 3rd party) and report them for local authorities. This could also reduce even fast food facilities and gas station misc. sales as the car could take you while you sleep meaning few stops of food and facilities.

vmulla | 15 juillet 2019

andy | July 15, 2019
I like the way this thread has developed - thanks for the ideas, they make good reading.

Nice, I believe in sharing and encouraging participation - you never know which idea will encourage a good discussion.

ReD eXiLe ms us | 20 juillet 2019

Lonestar10_1999: Before the Model S was released, Tesla tried to collaborate with the SAE, offering their own connector as a standard for both AC and DC vehicle charging. They were told to sit down, shut up, or get the [FLOCK] out, while people who worked for 'REAL' car companies did their jobs. So, Tesla left.

Collaboration, Cooperation, Cohesiveness certainly have their place. But apparently not among Competitors in the automotive industry. Ford and Mazda split. Chrysler and Mitsubishi broke up. Ford no longer owns Jaguar. The 'merger of equals' between Chrysler and Mercedes-Benz fell apart. The Nissan Renault Mitsubishi Alliance is in peril. FCA tried to find a 'new partner' in PSA, and that fell apart, to the relief of Chrysler, who felt that move left Detroit jobs on the chopping block.

Sure, Geely owns Jaguar/Land Rover now, but they largely leave them alone. Opel and Vauxhall were dumped by GM and turned around by PSA somehow. Generally speaking automotive manufacturers don't get along with each other very well. So, I don't have much hope that they'll be able to work together to come up with a workable standard for full autonomy. It would end up as convoluted as CSS has been.

howard | 20 juillet 2019

Yes, I look forward to not owning a car. When I need to go anywhere I'll call a RoboTaxi. This might be sooner in AZ than here in CO. Who knows maybe I can sell my Volt in the next few years. Turn the garage into a well-appointed workshop.

ReD eXiLe ms us | 20 juillet 2019

howard: It will have to be a very short 'few years'. Don't wait too long. The market for used ICE vehicles over ten years of age of any ilk, even hybrids or extended range plug-ins, is bound to fall off a 'demand cliff'... SOON.

"Twelve bucks a gallon! Please PAY before you PUMP!"

billtphotoman | 21 juillet 2019

This has been a great thread. I am fully aware just how dangerous driving a car really is. People talk about the deaths but if you want your eyes really opened look up the injury rates. They are nearly 100x as high (see https://aaafoundation.org/rates-motor-vehicle-crashes-injuries-deaths-re...)
Being in my late 50s and not wanting to be _required_ to drive once I hit may late 60s I eagerly await true level 5. At the same time having lived with EAP for nearly a year now I think it is farther away than most people think. The fact that enhanced summon which only operates the car at about 3 MPH line of sight is still not widely deployed reinforces this. But, I would really, really like to be wrong about the timeline.

billstanton | 21 juillet 2019

I concur with those supporting the thread. A good read. Thanks to Op.

Omicron I like your tying in the seat belts. Then there is air bags, then side curtain air bags. Then crumple zones. All resisted. "I would rather be thrown from the car than be trapped inside", "repair costs so much for new cars", etc. But bottom line they are safer.

It's hard to imagine how the FSD path will play out, nor if I would be happy losing my driving enjoyment, but I believe it will be a better future. Not too many other topics where I have that perspective.

mcmack15 | 21 juillet 2019

I was watching an interview with the Uber CEO this morning (think it was on Bloomberg). As for an opinion on timing from someone in the know, in the industry-------he mentioned two timelines. The 1st, in a few years (came across as sometime in the 3-5 year frame), he expected Uber cars to be doing 100% of the driving in VERY LIMITED situations. He explained that robots are good for routine and repetitive activities, of which both appear in about 1% of driving in a normal day. It is that 1% where he could see Uber cars doing 100% of the driving in that time frame. His second time frame was at least some 15 years out------where he could envision Uber cars doing what folks are hoping will be full driving.

calvin940 | 21 juillet 2019

"His second time frame was at least some 15 years out"

If theirs is that far out, they lose.

mcmack15 | 21 juillet 2019

I just watched portions of the interview again (I just discovered it was recorded earlier this month), and I misquoted above----he said the robots are good at 'repetitive' and 'predictable' (not 'routine' as I said above) behaviors. And he emphasized that robots and people will co-exist in this process. And the 5 and 15 year timelines are what he referenced----think he said at least 15 years.

Time will tell. It is a very interesting time we live in, and while the 15 year target will be beyond my expiration date, I look forward to seeing what will happen in the interim. I just wish I was smart enough to make some good money off whats coming to leave the grandchildren a nice bundle. It sounds like the auto industry is going to change dramatically------all these self driving cars and vehicles will cut way back on the need for personal cars; gas stations----I see them going away with the solar and batteries; auto supply stores; manufacturers of car parts, new and used, etc., etc.

I saw in the news tonight our buddy Elon has a contest going with some of the world's brightest young minds with his Space X company-------developing a train that can go to NY and Boston in something like 20 minutes in a tunnel!!!!!

ODWms | 22 juillet 2019

If it’s to happen at all, the 15 year time frame sounds more believable to me.

ReD eXiLe ms us | 22 juillet 2019

Seeing as Hydrogen fuel cell cars have been predicted to be 'about 15 years away' for the past 45 years, I don't feel comfortable with the guys who want me to wait even another 15 quarters, and I am far more trusting of someone trying to 'get there' within the next 15 months.

sky-pilot | 22 juillet 2019

And how will FSD work with a manual tranny? So for it to be the norm, they'd first have to outlaw manual transmissions. I agree with others, I don't see it happening anytime soon, at least not on a ubiquitous scale.

Magic 8 Ball | 22 juillet 2019

@sky-pilot The future is BEV and FSD, no "tranny" per se', in a BEV.

Pages