There are so many companies working on self driving technology now. All the tech used here looks really fancy and interesting. But I am worried that if one day people find out that commercial application of such technology is limited by the nature of driving itself, all hot money will leave this area(Given interestes rate is rising there will be less hot money in the future ). What's your opinion on this?
- Intel / EngD.J.TrumpmoreThis tech still got long way to go. I would not bet a computer to drive my car. All those radar and tech in the car will also cost a lot of money for the average joe to afford.
- @berrytree That couldn't be farther from the truth... You sound like a person who lives in Seattle and doesn't own a car yourself and has a jaded view of the 'average joe'. Over 76% of commuters in the United States travel alone to work in their own car. Your average car owner spends close to $500/month for their car payment less $100-$200 per month for insurance. 5 days a week, your monthly cost for a car-to-go service is about $900. That's IF you can even find an available car for your destination and source every day which would be aggravating to manage if you don't live in some down town area like Seattle. Something like Uber would cost you around $1700/month to use 5 days a week to commute for an average commuter distance.. So from a numbers and statistics perspective, not owning a car is not only less convenient, but it's also way more expensive if you consider the average joe..
- @B.bla That goes back to some other's points about infrastructure... Today's car-to-go services where you aren't paying for a driver are still more expensive than owning a car and there's very little density of these cars in suburban areas to make it a realistic option for your average joe. So yeah, theoretically, if you want to push for taxing the crap out of everyone to pay and have enough of these cars on the road to make it a realistic option then maybe it could be cheaper, but most people are likely to own a self driving car by the time that happens in my opinion.
- LinkedIn engdirA company like Waymo or Uber could buy *your* car from you, retrofit it, and charge you per mile to drive you to work every day. Other than the cost of the sensors (which is already coming down quickly), it could be cost-neutral to operate a self-driving taxi service without any car sharing. So it's already at least neutral with current car ownership costs.
* You could share them, even in very remote areas, occasionally at least
* They can do things like drop you off (valet parking) and get gas/charge on their own, saving you time
* You can do other stuff while it drives, like watch a movie or have a video call, adding value
* Shared cars can be much smaller (1-2 seater, not 4-5) because 90% or more of rides fit. You can order XL for the rare family trips.
* Smaller cars means cheaper gas/electricity, less road congestion, etc. Ends up being cheaper for that reason alone.
* Self-driven cars are, or at least will be, safer, meaning much cheaper insurance and repair costs.
So self-driving taxis are necessarily cheaper, in the long run, even for very rural areas.
- Amazon md jesusIt is the future. In trucking, running shuttles, buses and cabs. I still believe personal self driving car is still at a distant future. The issue is not the technology, it's rather the regulation and safety requirements.
- Oh come on, the issue is also the technology. How long have we had speech recognition (dragon naturally speaking anyone) and yet it still sucks. Regulation exists because not everyone has the same level of confidence as the typical yuppie overconfident software engineer. We’ve had autopilot for a while too- and even ones that can land the plane itself. How many people do you know that would get in a plane without a pilot?Feb 611
- New / R&DzorkanThe tech has been around for years. Self driving is not a particularly hard problem. There's a lot of startups in the space because most ML grad students can get a prototype working for their local area in just a few months.
The real challenge is productionizing and testing the cars at a global scale. Making it work in all regions and all possible scenarios is a big ask, and it will take time to get there in a safe manner. It's compounded by the fact that debugging ML systems is not a deterministic process -you can't just look at a stack trace and figure out what's going on when the car crashes.
- It’s all going great until you realize cars can’t drive in 1” of snow, ice, unpaved roads, streets which are not clearly marked, and currently can’t make many left-turns. All we need are perfectly paved and marked roads and an endless summer.
- In California it doesn’t snow regularly, people still like to hit the slopes. Self-driving car means you don’t have a steering wheel. Adaptive Cruise Control is nice, so is Lane Keep Assist, Collision Avoidance, etc, but it’s not self-driving now is it?
Tesla’s are overpriced death-traps. Anyone choosing a Tesla over an S-Klasse is an idiot!
- @jjpf13 BS. Self-driving has 5 internationally defined “levels”, the highest one is steering wheel *optional*. No one is expecting a current Model 3 to reach Level 5 let alone in all situations. Certainly no one is going to remove their steering wheel.
Okay, you’re clearly not a serious person.
@FBisEvil Intent to mislead? Do you have any idea how AutoPilot on airplanes works? No, it’s not a misnomer or any intent to mislead. It’s branding. Do you think any of the competitive systems (like Traffic Jam Pilot, etc) are intending to mislead as well? Why?
What claims of FSD are you talking about?
- Musk is overly optimistic and aggressive in his roadmap projections, that’s nothing new. That isn’t “fraud”. He should hedge his statements more maybe, but he does drive results.
Most of his statements, like the Recode one I believe, are actually “when will FSD buyers start getting something differentiated from EAP”. That’s very different from when FSD is “done”. It will come incrementally, and improve over time just as EAP and NavOnAP have.
Your attempts to spread FUD must be motivated by something, what might that be?
- FSD = Full Self Driving
Since undefined didn’t mention Musks actual quote and instead made up an excuse like a fan boi, I’ll provide a direct quote.
“You know, I think we’ll get to full self-driving next year. As a generalized solution, I think. But that’s a ... like, we’re on track to do that next year. So I don’t know. I don’t think anyone else is on tract to do it next year.
Of course being optimistic isn’t a crime, but intentionally misleading bond investors is. There’s plenty of other examples of musk lies, like model 3 ramp up rate, securities fraud which he is legally not allowed to deny (Occams razor anyone? Not for fan bois), and just recently in the last earnings call where he straight up lied about European regulatory approval to investors on an investor conference call. I was trying to stay on optic with the thread though.
I realize that musk acolytes will apologize and make up rationalizations, and that’s fine. I honestly don’t care. But we’re talking about self driving and I wanted to mention the rampant fraud in the company that’s allegedly was going to get it next year.
- Literally thousands of people die in car accidents every year and no one has made legislation banning people from driving. A federal ban on self driving cars will never have enough support to become law.
We haven’t visited mars because we’ve never properly funded a mars program, not because we don’t have the ability to do it. There’s plenty of funding for self driving cars.
- Google / EngMr GlassProblem with people dying when driving is that they're responsible for their death. People are far more comfortable having ownership over their lives despite being bad at protecting it, than they are with giving up control to another robot or entity no matter how much better they may be at safety. It's the classic utopian socialist vs free capitalist debate. People choose to be free (and poor) despite promises of Utopia (justice, wealth, safety) in socialism. The self driving car will be legislated and banned despite being safer for this reason.
- Elite that’s completely sensible but laws don’t follow logic when it comes to protecting people. If a lawmaker can win votes by outlawing self driving cars in their state, it’s going to happen. And people are scared of computers driving cars. There is already a backlash in PHX.
You can see similar problems with nuclear power today. Objectively it’s superior and safer than alternatives but that doesn’t matter.
- Salesforce marc2020!moreIn a litigious society like America, it will be really hard. It will take a long time to mature this technology. Imagine the combinatorial mess. Two self driving cars crash, one self driving crashes into a human driven car, a human driven car crashes into a self driving car, and so on. Whether conditions, mechanical fitness of the car (not enough friction in the tires or wiper in bad shape). It will be a paradise for lawyers and nightmare for ordinary people.
- Don’t let this thread make you forget that Tesla will fraudulently take customer deposits for Full Self Driver tech, despite the fact that those cars will be scrapped before that tech comes online. Musk claims it’ll come out next yr though.
Lol, what a subsidy-seeking truffle pig fraud boy Musk is.
- @RE5PECT that’s about the most ignorant comment you could make. Tesla’s stock performance is stellar and those that have performed better over the last 10, 5, 3, and 1 year are very rare.
@FBisEvil That’s like saying Kickstarter is fraudulently taking deposits. People pre-paying for FSD know the deal, and your claim that their cars will be scrapped before it comes is baseless. It’s a near certainty that those drivers will get increased AutoPilot capabilities for their money. Will they reach a full Level 5, no person in driver seat needed, works in all conditions capability? Almost certainly not. But will they reach a Level 4 (“mind off”) capability for some routes and conditions? Seems likely. They’re pretty darn close to a good Level 3 today even without their new upgraded computer.
Flagged by the community.
- Musk has claimed you’ll be able to go cross country without user input, and that your car will make money for you as an autonomous Uber while you sleep. That FSD claim is level 5 and he claims it is just on the horizon for years now. Recently he said it’s coming in 2019 in a November recode interview.
Try to rationalize it all you want but you sound uninformed of musks claims
- Amazon BuzzeringOP is ignorant. I don't blame him/her as it really depends on which side of the thought process you are on. FSD is far from realization, but for the money they were charging ($3000 against $5000 for Autopilot, and $3000 for a $100,000 car), it's an investment and they'll keep your car upgraded. For eg, with quantum computing coming to Tesla cars in Q2 2019, the folks who bought FSD (is not sold anymore) are receiving free upgrade.
Look, I honestly don't know when level 5 autonomous driving will come. But, the journey towards the goal is very exciting. And more importantly, it better start now.
- @undefined. I’m not lying here is a direct quote about full self-driving from the ReCode interview. You can google it.
“You know, I think we’ll get to full self-driving next year. As a generalized solution, I think. But that’s a ... like, we’re on track to do that next year. So I don’t know. I don’t think anyone else is on tract to do it next year.”
He’s done it countless times so quite trying to deny information available in the public space.
The 2016 interview on The Verge promised full self driving the following year. It was a lie. He doubles down on that claim all the time, recently in November 2018 in The Recode interview.
- Apple Yror78Of course it does. The answer to these kinds of questions is the same every time. It’s not IF it has a future, it’s what kind of future it will have....how far are we from maturity? Today, certain problems in autonomous drive are hard. Tomorrow, we’ll have solved these and tackle the next hardest questions. The answer is always somewhere in the middle with most things.
- What does Microsoft know that they aren’t sharing? Every Microsoft person in this thread is extremely pessimistic lol
- If you do the TCO math, you’ll realize a model 3 is cheaper to own than a Camry or accord over 5 years, and much cheaper than a comparable BMW.
Also the drivetrain has unprecedented reliability - google pictures of motor and gear that were driven a MILLION miles and still looked new.
There’s a lot of lack of education about Tesla (and EVs in general). If you actually drive one of these cars and can do basic TCO math, there’s no reason to consider any other car today.
- Airbnb BnglMyJnglhttps://comma.ai - self-driving cars won’t require well marked roads and will be able to drive on roads even covered in snow eventually as maps are built out and paired with highly accurate GPS then cars localize as they drive and use camera/radar/GPS/other sensors. Self-driving cars will happen, sooner than most of the above are guessing.
- IMO it’s a useless endeavor, never ever will see full potential on roads. But ... people working on it will be super stars tomorrow. Bits and pieces of their work will unlock many solutions in a few decades
- eBay flyingnowAll the technology are aiming at eliminating bottom workers. When will technology be invented to eliminate CEOs and VPs
This comment was deleted by original commenter.
- If you have driven a recent Tesla (HW2.5) and used Navigate on Autopilot, you’d know they are fairly autonomous already. If you drive from say Los Gatos to San Francisco, the Tesla EAP will get you nearly there with no intervention needed except for lane change confirmations (highway portion). And the need for those confirmations will go away in the next 1-2 months.
So highway full self driving is all but here ALREADY. Off highway is a bit tirickier and will likely take 3-5 years at most.
Again if you have doubts, rent an EAP equipped model 3 from Turo for a few days and experience for yourself.
- Having recruited ADAS engineers, yes this tech is here to stay. It will do as other technologies have done in the past, start as a premium option until costs come down and you can get it on a junk Kia.
We are in generation 0.5 more or less currently. It's a proof of concept that people are paying for to allow the companies to gather data.
Next will be refining algorithms and decision trees, refining it to mirror your own driving style instead of turning into "Sunday driver" mode.
The reason it will stay is that city planning will change. This will allow future cars to be parked further away from downtowns and city centers. Parking decks / lots take up a lot of space currently that could be turned into more profitable businesses or more green spaces could be made in downtowns if your car could drive up to you from a remote deck.
Additionally, you have people who are not car enthusiasts. The see cars as dishwashers, just a machine to get from A to B. They don't find pleasure in driving and will enjoy the time consuming content instead of driving. This will bring another revenue stream / partnerships within automotive once this technology has matured (like ABS and airbags before it).
The reason it has to be so complex is because of "shared roads." You have dumb cars with spark plugs and little else electronic sharing the road with the latest greatest. So the car has to see and "think" what to do in its surroundings. Due to lawsuits being as prevalent as they are in America, that's why you have such conservative autonomous cars today, companies don't want litigation IF a car does XYZ.
And that's my nickel tour TED talk on autonomous vehicles.
- Here is a perspective from someone in the industry: What we are forecasting is a 20 year transition to mass adoption and full penetration in mobility on demand fleets. Tesla and other legacy automakers will continue to rollout fancy partially automated cruise control systems, and these will become increasingly sophisticated. However the real first autonomous vehicles (aka - no "safety driver") will not be passenger cars. They will be low speed shuttles, cargo robots, etc. operating in geofenced districts or campuses. In the medium term expect waymo, zoox, cruise & others to roll out automated taxis and microtransit into select parts of cities come early 2020s. IMO: the real limitation is not the technology but rather transport planning and policy. Until those are sorted expect niche pockets of autonomy rolling out like a 19th century electrical grid in locations with the right traits, and localized and tailored to the specifics of their region. It's not like you flip a switch; the global car parc has a 12 year lifecycle.
- Have you ever driven in a fully autonomous vehicle? Can a Tesla drive itself on the highway or busy downtown streets without human supervision? I am not bashing their capability but simply pointing out the obvious. Tesla does not make autonomous vehicles. It makes great cars with useful safety features, but these features don't replace human driver and as implemented never will. Full autonomy is a very different problem to solve, and unlikely to happen on a consumer budget at least for quite some time. The sensing suite and compute in a full AV are on a wholly different level in terms of cost and capability. PS: I own a Tesla 😉
- This confusion always gets confused and overly complicated. If by autonomous u mean a car that can handle every single situation by itself we are definitely far from that. But for the most common use cases - parking, stop and go freeway traffic, etc the new tesla update is level 3 autonomous
- Red parrot - with Hw2.5 and NoA , the 3 can drive itself on the highway - see my post above. The S/X with older hardware may not be there yet though. It’s not perfect but getting better everyday. Human supervision is required by law and advisable but the system performs so well, I wouldn’t be surprised we get full high way autonomy within the year.
Off highway is a bit further out though.
- PS - the latest - https://www.inverse.com/article/52896-tesla-waymo-elon-musk-q1-earings-call-quotes
- Most people just talk about AI-aspect of it, but the underlying hardware is hard real-time system. Everyone tends to assume whatever software wants to do can be done by the physical system and the real problem is to figure out how to drive by AI. It is, however, actually not so easy to reliably control a fast-moving physcial system with hard real-time constraints by software. It would appear it works and follows whatever order it was given until one day some unpected case happens and the car crashes. Even if there's an algorithm that can make decisions to drive completely safely, the correct/perfect actuation of the car itself needs to be guaranteed and verified, which was never done really.
- im talking about a situation when traction control software detects the situation and wants to do something, but the unpredictable timing behavior of computer systems (this needs a phd to understand, though) delays the engagement and misses the critical time for action. how predictable do you think computers are, in terms of timing? indeed very slight chances of happening, but never there was a guarantee of preventing such misbehavior, and it is fundamentally unsolvable problem. (halting problem)
- From reading and listening to podcasts about this field, it is my understanding that none of the engineers, CEOs and VCs are expecting to make cars fully autonomous before launching to the market. Cars will have a human sitting to take over as well as have a remote operator, and also would most likely operate in a geo fenced area. There will be companies that would build the software, and then there would be car companies (like BMW) who would build cars that utilize these technologies. Something that will most likely happen in the next 5 years would be for instance BMW builds a car, Uber/Lyft provide self driven cab service using such a fleet of cars. Of course, like I said, to begin with they will be remotely operated and will have a backup driver. This is my understanding about their vision from listening to podcasts in this area. People who are working in this field are very optimistic because even before Uber became so popular, people were apprehensive about sitting in a stranger's car, and then it became so popular in a few years. I personally think autonomous cars have a future.
- DTE Energy / EngnhFO18Latency would be a huge problem on the road .
Remote piloting of military drones only works because they are not going to run into anything anytime soon . The Cars AI has to be able to handle every problem . I think the biggest problem self driving cars are going to have is unpredictable people driving there own cars next to them .
I drive in Detroit for 2 hours each day . And I spend a lot of time observing how driving works . One of the skills we develop over time is reading other cars/drivers “body language” to predict what they are going to do or are likely to do before they do it . Young drivers do not do this . They only looks for signals and break lights and then react. But experienced drivers and tell by the turn of a head or the slight movement of an car what they are thinking . I have teenagers, so I definitely noticed this while teaching them how to survive on the road.
- Yeah, I am not going to speculate on the details and edge cases. May be you are right. But I think they are most likely working on such problems. I am also not going to compare with military drones because I think sometimes silicon valley can be way ahead of the government. Facebook's face recognition vs FBI's for instance. I think a whole different talent pool is at work in Silicon Valley than in military. May be I am wrong. Also disagree that car's AI has to be able to handle every problem before autonomous cars can be launched. That's the whole point of backup driver and remote operator. Again, I don't have all the details so I am not going to speculate, but I think I am very hopeful and I think some of the best minds are at work on these problems.
- Define self driving.
Level 4? Where a car doesn't need a human in it ever, but can only do so within a heavily geomapped area like certain cities? We're only a couple years away from MULTIPLE companies getting that.
Cost parity of that with Uber/Lyft? Eh, more like five years probably, but by that point they'll drive better than most humans anyway
Level 5 is what you're probably thinking of, completely autonomous absolutely anywhere on planet Earth. That's at least a decade away but also not that needed. Level 4 will enact massive societal change, level 5 will be a neat party trick by comparison.
You aren't going to recognize transportation in 2030
- OpenTable MeliodasTrains don’t even have the tech yet and all they do is go in a straight line.
- 1-800 Contacts HQOR61There’s such a huge incentive to perfect the technology, that people will keep chipping away at it until it works. Not having to pay truck drivers will save billions and billions of dollars. It will be bigger than Amazon.
- The first autonomous vehicles will succeed in situations where the built environment develops to meet them in the middle.
We assume that the vehicles will have to be brilliant while the streets remain dumb. Instead, there are plenty of situations where adding a modicum of smarts to the environment will make it much easier to hand control over to the system.
We have carpool lanes now- autodrive lanes with embedded sensors and transponders, fixed cameras with machine vision etc are a possibility and will let you take a nap in your otherwise level 3 or 4 car.
Municipalities will be incentivized to fund this infrastructure as it will be many times cheaper than all new roads or mass transit.
- Virgin Hyperloop One JVew62I’m going to go one-step above this; transportation is an issue. I live in LA. Getting from point A to B can be a hassle. This is true for most people in densely populated areas. Self-driving cars are a step towards having efficient traffic. Humans make errors and poor judgement calls when driving. These choices can lead to bottlenecks, accidents, etc. Self-driving cars address this. Picture the ideal situation with roads that have only autonomous vehicles—traffic could be fully managed and optimized. The challenge is working in the grey area before that...with humans in the loop. This is where we are now and will be for a while. Autonomous cars offer a solution to a problem that is going nowhere. Transportation is an issue. Whether it’s addressed with self-driving cars, scooters, flying taxis, hyperloops, high-speed rails, other infrastructure, or whatever, the problem needs solving.
- As much ad I would love to be driven around by a piece of code, I would probably be scared shitless on the highway. Hard to give up control.
- Weird as it sounds, highway driving is actually way easier to automate than busy city streets. It's easier to calculate kinematic interactions since everything has crazy inertia.
Also an av will hit the brakes way harder than you will, humans tend to be really cautious with the brake pedal and only hard brake when they realize oh shit I'm not braking hard enough. An av will be like "that swerving car is X far ahead, so I need to brake exactly Y hard not to hit it"Feb 111
- Think about how many different car brands are around. There will be consolidation as a google and auto companies gobble smaller tech companies. The risk is if the experience gets commoditized. Also different types of vehicles such as trucks, mini bus, buses and other mass transit options will create additional options.
- There will not be flying cars that’s not viable for so many reasons. Google may be involved in self driving cars don’t doubt that but don’t think they can dominate the consumer space. Tesla still has a massive edge and it’s great all the people betting against them. I think musks vision for autonomy and also underground tunnels will prevail
- I don‘t trust and even believe in self driving cars! This is one of the things the world really doesn’t need. If you want a driver take a cab or bus. It‘s that simple
- Salesforce hfgnntunvUntil there is AI (real one not marketing gimmicks) or all human drivers removed from the roads, self-driving will remain a novelty
- Google nooglr101L4 should be ready by 2025, at lest in limited scope (city, time, weather etc). L5 - no one knows.
Question is which stock to own when L4 go main stream in a decade?
- Indeed / OtherTC1BillimoreSeems as though the finest brains working on this non this just can’t figure out edge cases. The edge cases are how people die. So who knows. It seems as though generally 95 % maybe has been figured out maybe even higher but the edge cases are what make it nearly impossible to get to the level of “full self driving”. AI/ML these day seems to be more feeding massive data sets to algorithms to decide best decision but humans inherently enter their biases in these datasets. Do we really have true intelligence? Who knows I wish I understood more about the field tbh.
- What are the edge cases? The self driving car should always follow the rules of the road. It shouldn't have to decide if a bus full of Rhodes scholar teens or itself needs to be saved, just follow the rules of the road and crash avoidance rules and the bus driver can fend for themselves.
- Nvidia Biqk71I think of achieving full autonomy for self driving vehicles as another landing on the moon project. You will need a county willing to build the required infrastructure and a partnership with industry. May be the US will lag behind other countries that are more determined to achieve this.
- Chase SamanthaSNO. Then what is the point of humans be alive. Self driving cars again need fossil fuels unless they are battery charged. It will take it's own time may be close to 50 years to evolve . But self driving cars like in movies I will prefer a car that I drive and uses electric. And does not make me lazy. Humans are already lazy life expectancy is reducing. So be with it. Eat sleep sit in self driving car . What a life. 🤷🏻♂️
- Optum / ITzkzkzkzkIt will require an infrastructure change also. Just like some airplanes can self land on certain runways with ILS categories. Certain road will be AP certified for full self driving where humans shouldn’t be found walking for example. Eventually getting more and more roads certified until the system is fully vetted.
That’s my contention.