Waymo to customers: “Completely driverless Waymo cars are on the way”
techcrunch.com
techcrunch.com
https://www.theverge.com/2017/11/7/16615290/waymo-self-drivi...
Hopefully this time it's the real deal. I remain skeptical.
This announcement is just a repeat of what they've said before. If they actually do it this time then that's great, but they haven't actually said anything new here.
Sorry you felt that was disingenuous! You're right, it's not public in that anyone can take a ride, your account has to be approved.
This just reinforces my point. Waymo can announce anything with vague terminology of things happening "soon" but we still don't have any verification of anything happening.
Note that while that risk remains true of every transportation journey you take, there is a lot of press generated when an autonomous vehicle takes a life or causes injury, which plays a part in a lot of peoples decision making.
Also note recent articles about automatic stop detection failing >50% of the time. This may or may not be relevant to AVs, but I personally would definitely not want to be in an AV that hit a child. That (theoretical) child's life is worth more than pushing driverless cars into the world before they're potentially ready.
Waymo is still Waymo whether they're running rides on their own behalf, or under the Lyft name. Someone who's ridden Waymo in an early adopter program may make their decision on whether to ride Waymo under Lyft based on that experience.
Get it?
FTA:
> "are on the way"
> "soon you may experience"
Vague promises without any concrete dates attached make this seem like output from PR rather than engineering
> since mid-October, the company has been operating its autonomous minivans on public roads in Arizona without a safety driver
The difference is that those were without any users inside. This new article is saying that the public offering, with real customers inside, will get rides with no one inside.
It is unfortunate that it took 2 years to get to the point where they felt safe putting people customers in the car, but it's very misleading to equate the two and use is as a way to imply they lied last time around.
Except they're not at that point yet, and even when they are they will still have a Waymo employee on board "initially".
I don't think that anybody is accusing anybody of lying, just making the point that predictions of the marketing team aren't really something you would want to set your watch by.
I think you're confused. They've been having customer rides with employee in the car for almost a year now. The email sent to customers today is saying that they will soon find rides with no employee inside.
And again, rides without employee inside have been happening also since 2017, just not with customers inside.
The point is that "soon" is meaningless in this context and Waymo very clearly states that "a Waymo employee would likely be present in the vehicle initially."
1. Waymo has been having rides with no one in the car for 2 years now.
2. Waymo has been having rides with real customers inside for 1 year now.
3. Waymo has told said customers in #2 that they may soon see rides with no employee inside
I agree that the soon in #3 doesn't mean much, but at the same time, this wasn't a press call, it was a private email sent to users, so it's not like they're showing off.
But 1 and 2 have already been happening for a long time.
Of course. Companies tell their customers about new products in the hope that they will keep that info secret.
I think the salient point about your list is that while 1 and 2 have been done separately for some time, the only timeline we have for them happening concurrently is "soon"
https://www.wired.com/story/designated-driver-teleoperations...
> But those cars won’t be operating completely unmanned — at least for now. Under these regulations, driverless cars being tested on public roads must have a remote operator monitoring the car, ready to take over as needed. That remote operator — who will be overseeing the car from a location outside of the car — must also be able to communicate with law enforcement as well as the passengers in the event of an accident.
That being said, the way Waymo does it is different from other companies. As HN can understand, adding the ability to remote access a car is very very risky. Waymo on the other hand has the ability to suggest a path to the car, and the car will itself locally evaluate to see if the suggested path is safe before executing it. So it's not quite a "remote control" access.
[0] https://www.vox.com/2018/2/26/17053898/driverless-cars-self-...
If the goal is to have the car be able to ask for help when it's confused though, it would likely be fine.
Your belief that the GP thinks they lied may be rooted in your own biases than in her or his words.
1) Say what but don't say when.
2) Say when but don't say what.
Never, ever, ever bet tempted to say something specific will happen in a bounded time range.
Flying cars are on the way! That disease you hate, a cure is coming! Next year will be a big year for other breakthroughs not mentioned here.
If you can't keep promises, and don't want to bullshit, then talk more about what the product already does.
OR
I can have many small sub-features that relate to this major feature done in 2 months, but I might know exactly which features make the cut in those 2 months.
Seems congruous to me
I agree and it's such a golden rule that no one follows or cares about even the slightest. Features and deadlines are to SE like velocities and positions are to QM.
This idea more thoroughly described here: https://news.ycombinator.com/item?id=21212087
But then again he probably uses deadlines as both carrots and whips.
Also seems makes good PR and gets people excited, even when exaggeration is accounted for.
The other side of me is absolutely not ready to trust my life to a machine - even though I know I do that all the time in other ways in modern life.
I think it will be people's emotional evaluation that will matter more than anything else and it will be a bumpy road to acceptance.
Driverless cara will probably be safer statistically but they’ll simultaneously make errors a human wouldn’t, so some people are going to die in what will seem like really dumb ways. This will be hard to accept even though that already happens with human drivers. Probably because their mistakes will be easier to identify with and explain.
Like, I'm cruising along at 45mph and a plastic bag is blown in front of my car. Yes, I see an obstacle and need to make a decision about what to do - but I do this as a human who knows what a plastic bag is, and how it is far safer to drive right over it than slam on the brakes. I know a plastic bag is not, say, a Pomeranian.
Another example: I'm driving, and from beyond my plane of vision, a basketball bounces onto the road up ahead. I can't see past a building, but I know I'm near a park where kids frequently play, and I know it's possible a kid comes running into the road to grab the ball. I proactively slow down. I guess you could say a leavel-5 autonomous car would see the ball and slow down anyway, but there's still that lack of cognition that concerns me.
We like to think of driving as a purely rule-based game that is simple enough to model and train against. And I believe in 99% of situations, it is. But in that 1%... the plastic bags and all that...
But at the same time that AI driver will have stopped in time when 10 other kids jumped out without any warning because it has much, much better reflexes.
But it’ll be the first instance that will get all the airtime.
Think about that.
Self driving cars aren't just a solution to nuisance of driving. They're a solution to one of the leading causes of DEATH and INJURY in the country.
Sure there are going to be edge cases where we as humans will be able to point at the car and say "damnit, we could have prevented that"!
But I sure hope when we do that, the car points right back at us and asks "And for the other 1.20 million lives I've saved this year? You would have killed them"
This is a meaningless assertion unless you account for how often "edge cases" arise. If unpredictable pedestrians, unplanned road construction, etc., is an "edge case" the car basically won't work in DC or New York.
Sleepy truck drivers are a major safety issue by themselves, and highways are easier for SDCs to drive on.
no, they will, at some point in the future, be a solution. until we seen serious miles in comparable conditions they are an unproven technology, that why I appreciate waymo approach of getting from here to there slowly and carefully compared to Tesla which has a dozen death already under the autopilot and hides them under a pile of pr and terms of services.
but for all the tech they throw at it until the fundamental issues of cognition and prediction are solved they will only be able to reduce a fraction of the accidents because they aren't replacing the most dangerous vehicles on the road and they only are better for a subset of cases, like inattentive or impaired/intoxicated driver.
Sure, humans drive a lot. The idea is that self driving cars will also, drive a lot.
Nobody is saying humans are terrible drivers (well, some maybe are). But if there's something better...
If reducing DEATH and INJURY are the motivation for self-driving cars, we can achieve a larger reduction with technology that is available right now: public transport. And there is even a no-technology solution: a drastic reduction of road traffic that passes through residential areas.
Putting more cars on the road might or might not reduce DEATH and INJURY, given that those cars are going to be self-driving. But _removing_ cars from roads will _certainly_ reduce DEATH and INJURY- and also reduce the environmental damage from the cars' internal combustion engines and the infrastructure needed for those cars to circulate.
Considering all of the above, I'm really not convinced that the self-driving car industry is really that interested in reducing DEATH and INJURY. I think their main motive is to increase PROFIT.
That said, the benefits of self driving cars aren't precluded by the motives for developing them.
While I hear your points about removing cars from the road, and generally agree with them, there are definite benefits to cars which are arguably not replaced by things like public transit. There will always be a need for cars, so lets make them as safe as possible. If self driving is a step in that direction, sign me up.
If AI can't recognize some obstacles at all, then I think it's safe to assume there may be some delay in obstacle recognition compared to humans.
1) Everyone assumes the good will outweigh the bad. (And why proof of humans doing one task better than AI is somehow also taken as self-evident proof that the AI will be better at everything else.)
2) No one assumes that we, as humans, will ever be able to improve our driving skills in way that can keep up with AI advances. Someone could invent a magic driving pill. Which is crazy. I just don't assume that it's any crazier than a near-perfect driving AI.
An AI driver would have 360 degree coverage with much faster reaction time - think about the cases where a car brakes before the passenger even realises there is danger ahead.
> Using a child-sized dummy, vehicles only avoid running over the child 11% of the time
They're not safe.
Not the same thing at all.
I don't think that bodes well at all for full autonomy - especially since at least one of those vendors (Tesla) is promising self-driving sooner than later.
I would argue that the examples you gave are just image recognition, etc. What is your brain doing if not image recognition and obstacle avoidance?
We're not there with today's technology but eventually our machines will perform better at those uniquely human tasks you mentioned. After all, we humans and our cognition is just the output of these giant meat computers inside our skulls pulling in sensory data and running pattern recognition algorithms.
The $65,000 question is whether or not today's technology is sufficient (however barbaric) to be acceptable for the job.
Planning, anticipation, reading body language, etc? Is that person looking like he’s going to jaywalk? That’s not image recognition. Yes, once he steps into the road the self driving car will respond quicker. But the human driver might already be slowing down and watching for hints of initiating movement. Image recognition is just one tiny piece of it.
How do you do it? Telepathically? You're still just receiving visual input and processing it.
Knowing there is a location ahead without a crosswalk that kids constantly cross at.
Or knowing that the car in front of you is going too fast, implying that they're unaware of the railroad tracks they're about to jump.
You don't need telepathy to make useful predictions. And humans don't need to be accurate, they just need to be prepared to take a measured response.
That's not based on reasoning. Most people driving by probably don't even know. And it might be in the car's database.
> Or knowing that the car in front of you is going too fast, implying that they're unaware of the railroad tracks they're about to jump.
If I'm behind them, and I'm going much slower because I know the layout of the road... I'm not sure how knowing this matters?
Tesla track where every Tesla goes, all the time. Will Apple make a car before Tesla makes a 'smartphone'?
There are still a nearly endless number of things the car can't get from on-board sensors, but that people will know, or at least be able to learn.
For children you might preemptively slow down, but a properly designed system can brake a whole lot faster. Should be fine.
1) If I'm considering whether to brake for a plastic bag, I'm also going to check for anyone tailgating me, and take that into consideration. Including whether said tailgater is a geo prism or a Freightliner attached to a tanker of gasoline.
Slamming on the brakes for a plastic bag is not the universal safer choice.
2) Already mentioned this, but reaction/breaking time is meaningless without timely recognition.
...I didn't say it was. And you call me the one making assumptions.
> reaction/breaking time is meaningless without timely recognition.
If it can't quickly recognize large moving objects then it's not qualified to drive at all. I'm assuming it can do that because self-driving at all is part of the premise.
(Well, specifically the recognition is necessary for basic driving functionality, and being faster is a matter of adding compute power and using cameras at least as fast as the ones in an iphone.)
2) >>>you might preemptively slow down, but a properly designed system can brake a whole lot faster.
>If it can't quickly recognize large moving objects
I'm not questioning whether it can recognize quickly. Just whether it can ever be better/ faster than humans, taking into account cognitive ability.
It's whether human
(prediction-or-recognition time + reaction time + breaking time)
is more or less than
(AI recognition time + plus reaction time + breaking time)
If AI can react and brake 2 seconds faster than a human, but doesn't recognize the threat until 4 seconds after humans can, then AI drivers will kill more people than human drivers. Which won't be fine.
>being faster is a matter of adding compute power and using cameras
There are other comments about the limitations of relying only on image recognition. While compute power per might improve the AI recognition time, humans still have the advantage of being able to predict instead of react. This will continue to be a major limitation for AI.
Okay, but I didn't say that either. I said it would be an improvement to the experience I have. I will trade an abrupt stop in front of a plastic bag for several hours of being able to do what I want while the car does the driving.
> If AI can react and brake 2 seconds faster than a human, but doesn't recognize the threat until 4 seconds after humans can, then AI drivers will kill more people than human drivers. Which won't be fine.
The AI processing time to recognize things in a frame needs to match the framerate, or the system is just broken. Maybe it waits for 3 frames to show the same thing before acting, that's fine. That's still faster than a human. Especially if you use a 240Hz camera, which is not very expensive. And you can and should put in enough computing power to run the 'reacting' system at the same frequency.
Then once it decides to brake, instead of waiting for a foot to move it can put the virtual brake pedal at any position in under a millisecond.
I can imagine a car so poorly designed that its reactions are slower than a human. But it's hard to imagine a car that's simultaneously certified for letting me take my eyes off the road, and is that broken.
So I still say that a self-driving car would be able to brake quicker and harder, and it should be fine. Even without the ability to predict.
One word: Chauffeur
>The AI processing time to recognize things in a frame needs to match the framerate
If you're talking about framerates at all, you're still missing the point. I'm talking about human ability to predict and anticipate. The fact that a human can brake before there's anything in the frame to even detect. We've (humans) got a head start, so the speed of the AI isn't all that important.
But for people with normal incomes too!
> brake before there's anything in the frame to even detect
The scenario was slowing down a bit, but still needing to hit the brake in response to the child. So if the AI can brake faster it is extremely important.
Up to a point - ordinary seatbelts even with smart pre-tensioners aren't great at this, and it may increase the rate of whiplash injuries for the passengers who aren't expecting braking.
We really should do this like the Dutch: when something problematic is encountered, engineer and regulate the hell out of it. Not just for self driving cars, but it works for humans to some extent that the laws are actually enforced.
Waymo cars have driven over 10 million miles so far without any fatal accidents. That's not nearly enough miles to compare against human drivers, but it'll be interesting to see who does better once they log some more miles.
You need a certain hysteresis/threshold to filter out false positives.
This isn't just a problem for driverless cars, it's become a real problem for people, too.
The rules of the road are there for a reason, people!
When you wave someone into traffic, or have your emergency flashers on while still moving in the rain, or wait too long at a busy 4-way stop, you're not being courteous or defensive or cautious, you're being unpredictable! (And most likely inefficient, as well.)
Worse, (and this is going to be driverless cars' Achilles heel, IMHO) is people pulling into or across traffic (or otherwise failing to yield) based only on the expectation that the vehicle with the right-of-way will brake and do whatever is required to avoid a wreck.
If they do it in front of a driverless car while trying to cross a busy highway, then people making this assumption are going to get killed.
Tesla’s last automation talk was interesting about that - they’re basically training the cars to act like a human driver would, looking for tiny signals that might warn of unexpected behaviour.
And it could have both problems. There's no reason I can think of to assume mutual exclusion.
>act like a human driver
Anything AI making this or a similar claim is automatically suspect to me.
>looking for tiny signals that might warn of unexpected behaviour
I suppose this is progress, except that it sort of implies that the AI wasn't already doing that. Which to me says the programmers are still having to anticipate too many possible scenarios.
Humans have an uncanny ability in predicting the overall movement before any perceivable deviation. Of course I'm not saying it's magic; there's obviously some kind of signal to our subconscious that we're able to process and make useful. Perhaps AI can pick up on some of that.
But it's not necessarily the "first movement" of the swerve that is most useful. And the "signal" that humans pick up on may not even be something externally measurable by the car. Like knowing of a particular interstate exit that drivers always seem to almost miss, but swerve across lanes to catch at the last second.
There are also signals that are externally measurable, but have to be learned (seen on a previous trip, or warned by a friend). Like knowing about a bad pothole ahead, and knowing that the car in front of you is going to turn the corner, slam on the brakes and then swerve to avoid it.
These are the kinds of things that I suspect AI will struggle to keep up with in the foreseeable future. And humans rely on this type of intuition way more than I think most people realize.
Of course I'm not saying it's magic; there's obviously some kind of signal to our subconscious that we're able to process and make useful.
That signal is connected to some much larger dataset of years of human experience - things like what a distracted driver might do, what a tired driver might do, what an old unmaintained vehicle looks like, what hints at an out-of-state or out-of-country driver who might not be aware of local routes or problems, how people tend to behave differently at dawn/rush hour/day/night.
An AI car should be able to tell which is a busy junction pretty easily, but watching a car swerve in a way that might indicate the driver is not paying full attention will make you more cautious around that car for a long time up ahead.
Still, AI cars which don't tailgate, will get an ENORMOUS safety advantage over the people who drive 2ft behind me at highway speeds. That alone might be enough to overcome every other benefit. "AI car doesn't get involved in multi-car pileup", "AI car didn't get shunted at traffic lights", "AI car let me in at that junction" will keep happening and keep boosting people's confidence in them little by little.
Yes, it's good to know this has a name. Thanks.
This idea, but also extended to include the fact that as AI drives more, humans drive less, and might lose some of that knowledge. If AI is driving 80% if the time, but can't handle 20% of situations, then humans are doing the 20% most difficult driving with 80% less practice. Obviously this is over simplifying things, as lack of fatigue and other things will come into play. But I'm merely stating that both sides of the equation will likely change, not just one.
>Still, AI cars which don't tailgate...ENORMOUS safety advantage
As long as they do this in a way that doesn't create congestion, I agree. But I'm not yet convinced. Overly conservative driving can be just as, if not more dangerous than some aggressive driving.
A lot of the risk might be merely transferred, not reduced. Preventable congestion is the bigger problem, tailgating is usually just a symptom. (Not saying it's ok, just that it's preventable in other, more indirect ways.)
And many solutions to congestion require complicated risk-now vs risk-later evaluations (including some that you listed- id'ing a bad driver or a tire about to blow out). These require more decisive, if not aggressive, driver actions, like speeding for 20 seconds to get away from the bad driver or to clear the passing lane. Or accelerating aggressively on an on-ramp (and not have to stop at the end of it to wait for an opening, which is part of why on-ramp metering has become so popular).
Or if you're on the interstate and it's posted for 45mph, but everyone else is driving 80mph. It's probably safer for you to speed (at least 60+) than to drive 45mph.
That's not really compatible with AI's current approach of being overly conservative with everything.
All this to say, I don't doubt AI will make some aspects of driving safer, but I still have doubts about whether it will improve overall safety, at least in the foreseeable future. Driving and traffic systems are complex.
That's something machine learning can do pretty well, actually. There has been a lot of research on this topic, and not applied just to self driving - natural language generation, stock prediction, playing games with multiple agents, even content recommendation do future prediction based on past information.
I'm not familiar with the games, but from what I've seen with all the other things listed, that's not a technology that I'm anywhere near ready to trust my life with.
It's obviously the best choice to adapt a technology like that, but I'm not sure our society has the fortitude to do it.
Seriously. Why does everyone assume this? Because Tesla's marketing department says so?
It would of course introduce a class of errors (B) that humans would not have made, but those can be iteratively improved until they're substantially lower than the category they're replacing.
My point is that once B is substantially lower than A, the tech should be deployed. But I fear people will demand B = 0.
>once B is substantially lower than A
It's the presumption that this ^ is a given; that's what I have a problem with.
I think people assume B will continue to decrease in a linear fashion, (or maybe log/exp, but at least not hitting it's asymptote / limit until it's less than A).
But we're already seeing diminishing returns, and since A and B are both difficult to measure, I think presuming that the good will automatically outweigh the bad (at least in the foreseeable future) is nothing more than blind hope at this point.
There's also no reason to believe that it won't become more dangerous at some point.
Look at air travel. Automation made great improvements in safety without many downsides. Until maybe about 10 years ago, when automation related accidents started creeping up (yet usually blamed on pilot error). And then you get to the 737-max problem.
(I'm not claiming air travel is more dangerous, it's not. But mechanical reliability and materials testing / repair are doing far more for safety than automation is. Modern jet engines are an underappreciated marvel of science.)
Personally, I would accept a greater risk for the convenience of being driven by AI, and I don't think I'm alone.
I just think it's a bit of blind hope (and even some arrogance) to assume AI will be safer anytime soon.
driverless cars will save thousands, and kill hundreds
Statistically it would take tens or hundreds of years before we can know whether self-driving cars are safer than human-driven cars:
Key Findings
∙ Autonomous vehicles would have to be driven hundreds of millions of miles
and sometimes hundreds of billions of miles to demonstrate their reliability
in terms of fatalities and injuries.
∙ Under even aggressive testing assumptions, existing fleets would take tens
and sometimes hundreds of years to drive these miles — an impossible
proposition if the aim is to demonstrate their performance prior to
releasing them on the roads for consumer use.
∙ Therefore, at least for fatalities and injuries, test-driving alone cannot
provide sufficient evidence for demonstrating autonomous vehicle safety.
Driving to Safety
How Many Miles of Driving Would It Take to Demonstrate Autonomous Vehicle Reliability?
~50 people died when a compounding pharmacy had mold in its spinal injections. ~300 died in SK when a ferry sank. Over a hundred people died due to Boeing's software issues. Hell, look at Uber killing that single pedestrian in Arizona -- how many traffic fatalities occur during Uber rides overall? This is all since 2010 with "advanced" technologies. With scale of tech, comes fringe/rare events, which will be reported on.
The important question is: will less people die with Waymo cars than without them? All logic and reason point towards "less people will die," but evidence is needed.
The one thing we can be sure of: when people DO die with SDCs, it will be HEAVILY reported on, even though thousands die daily on the roads due to preventable, human error, which is rarely reported on.
I've also seen Waymos make panicked Lane changes near the JCC many times, willing to come to a dead stop in the right lane and wait indefinitely for a break, even though it's an easy go-down-another-block-and-Uturn
I don't think you're being very accurate there. There are lots of videos of people commuting 50+ miles to and from work every day without touching anything while on the interstates, including interchanges. It's doing a lot more than just lane keeping.
Perhaps you should check more frequently? Here is a video of a Tesla on Autopilot stopping for a stop sign:
https://electrek.co/2019/05/29/tesla-autopilot-stop-sign-mak...
> “The car always stops at the end of a highway exit since Navigate on Autopilot. At the stop, you just need to touch the accelerator to start Autopilot again without Navigate on Autopilot. I didn’t accelerate, you just need to touch the accelerator for less than a second and the car will do the rest at the intersection.”
[1] https://www.teslarati.com/tesla-autopilot-fsd-traffic-light-...
Which cars have you drive with a superior cruise control implementation?
I feel like people forget human drivers hit those same barriers with alarming frequency.
That being said, Tesla's cruise control capabilities are best in class, but it's misleading marketing-speak to call it true self-driving.
What I mean is, you are not providing input to the wheel, or pedals, or turn signal or driving the car in any way. You're just letting it know you have not fallen asleep.
And, yes, thanks for asking, I have used it.
The problem with the latter approach is the long tail of real life variety doesn't lend itself to a fixed set of programmed rules. You'll always have more edge cases.
I think Waymos approach gets you to 90% faster, but then plateaus.
It is already full self driving in some cities, whereas my Tesla cannot even detect stop signs.
I love my Tesla but I harbor no illusions about the eventual victor of this self-driving race.
1:49:30. I’m on mobile otherwise I would link to the exact spot.
Tesla has tens (hundreds?) of thousands of vehicles on the road right now that are continuously feeding training data back to Tesla. What evidence is there that this is insufficient, or even deficient? From what I can gather, Waymo has limited testing geofenced to Phoenix, Arizona. Tesla has cars all over the world with autopilot, all providing training data.
I don’t see a reason for your conviction that Waymo has an advantage here.
So you have a lot more Teslas sending a lot less data each. I'm not sure which company has the richer training set.
Are they?
- they could store data to be collected in bulk when the car is in the garage for maintenance or servicing or repair.
- Superchargers could include fast data transfer in their cable for cars using them.
- Superchargers could include a wireless data collection from cars sopped at or near them.
- Cars could connect to public wifi to upload data at any car park, any place, any time.
- Tesla could have vehicles driving around, like Google mapping cars, collecting data wirelessly from nearby Teslas.
I don't know if they do any of them, but none are unreasonable things for companies of 2019 to think about doing.
I honestly don't expect go-anywhere L5 in the next decade at least, and I definitely expect an engineered approach to hit L4 first.
I'm not sure how long it will take, but I'm sure we'll get there.
I suspect that isn't true in so far that engineers will build what works. Building a complicated system like an autonomous car will necessarily involve a messy combination of a bunch of different strategies and with the result converging on a common solution.
Honestly, if your only critiques are based on hearsay then maybe you shouldn't critique.
I’m going to ignore the fact that you’re flatly incorrect about the rest of your posts regarding Tesla, since others have more ably corrected you than I could, and simply ask what the hell this even means.
A “production revenue service”? This sounds like meaningless babble. Waymo does not currently have anything available that you can buy, lease, subscribe to, or otherwise exchange money for. That is, in fact, the entire point of the article: “coming soon,” they say, with an additional clause on top of even that: “coming soon you may...”
So: what do you mean when you say “Waymo is in production revenue service”? Because that sounds like nonsense, to be honest.
We have lots of errors and junk DNA. We make all kinds of mistakes. Perhaps not fatal ones yet, we'll see how climate change plays out.
As for our perception and reflexes... I wouldn't brag that they're that good just because they haven't been bad enough to cause our extinction.
That's my point. The natural selection process for self-driving automobiles is just beginning, as they meet the challenge of the real world. I'm concerned that it is going to mirror the messiness that biological evolution involves.
But computers can run 10,000 simulations to learn in minutes. And obviously we are not going to create a autonomous driving ability in minutes but we can speed up the learning evolution significantly in machine world vs humans that have 10-50 years per evolutionary learning event.
My view is the hard thing about autonomous driving will be the change in style of accidents. Even if autonomous driving has half of human accidents we will probably see things that to the human eye seem really obvious so people will be 'if it can't predict that why should I trust it' type thing. But we won't see the many accidents a computer avoids that a human would make or finds obvious.
I'm positive about it but suspect we will see the roll out take many decades as it starts in cities with the right conditions, moves to trucking routes but will probably take some time for national metro coverage.
Point stands. I computer will do millions of simulations in a day. And be done across a population far greater than historic evolution.
Doesn't take long to get billions of learning events. On top of which evolution pre-intelligence (as a very loose term) was somewhat random, where our systems are starting from a much more evolved position that does take advantage of existing evolution as much as humans do.
There's a reason they're testing in Phoenix. It has almost no Inclement weather. Throw a little snow, mud, or sleet in there and it gets hairy fast. Level 2 is basically the 20% of effort in 80/20. It's exponential effort from then on out.
When autonomous drivers have the same or fewer injuries per billion miles driven as human drivers, then for all practical purposes it doesn't matter if it's level 4 or level 5.
If so why go through all this trouble to creat autonomous vehicles? I, for one, am glad this was not the prevailing attitude when commercial airliner regulatory bodies were set up.
1. They're cheaper. This sounds crude, until you realize that this makes things like buses and shared rides far more viable. Meaning millions of people have a better commute and collectively save many lifetimes worth of time. It can also mean less air pollution, due to less traffic congestion (this only applies if congestion pricing is done), which will also save lives.
2. They can always get better over time. Humans can't.
3. Fewer parking lots are needed, which makes cities far more pleasant to live in.
4. Disabled, elderly people and children have better mobility, and aren't dependent on others to drive them everywhere.
IMO autonomous vehicles are just going to further push public transit into something just for the poor. All the convenience and comfort and privacy of sitting in your own vehicle, but now you don't have to drive yourself! Why would you choose anything else?
buses would be cheaper to run if they didn't need a driver
So, yes savings. Nyah on your "unions defending humans against exploitation are the worst thing ever" narrative.
Also, if you don't have to accommodate a human driver and controls, many safety features, bulky gasoline tank, and trunk space, automobile design could change radically. Self-driving taxicabs that look like small pods, or covered golf carts, could be used for individual travel, increasing space efficiency.
So we can sleep in the car. Or watch netflix. Or drink. Saving 1+ million per year while you know, very nice, is not the main reason for doing this.
I think the future of self driving vehicles will just be inclement weather delays like we have with flying.
One of the lessons of engineering history that self-driving car proponents fail to grasp is that promising technology often fails to get traction because it can't close the final 5%. Supersonic aircraft, for example, have effectively been derailed by people on the ground near airports complaining about sonic booms. VTOL aircraft never replaced aircraft because of control problems. Flying wing aircraft never became popular for similar reasons.
Also, the prospect of being accidentally killed in a horrible car crash is less poignant if the alternative is more time spent in Chandler.
If markets are competitive the profits will be lower. If markets are not competitive (not many companies can make or license self-driving AI) the rest will sell cars with steering wheels.
The variable that is actually important is demand. If lots of people want their own car, or want to drive their car, the market will provide those things. If most people are content to rent, ownership might become an unusual luxury.
If for some reason charging technology is unable to be improved to the point that it's comparable in time to filling up, and if for some reason the solution is to go back to the idea of battery swaps. I could see this being reason enough to keep the whole system as rentals.
If I'm an owner, I'm going to think twice about swapping out a part that might be 50% of my cars value.
There would need to be some serious quality assurances to make that work.
New lidar was spotted a few months ago and more recently on the Jaguar cars and the CEO said it’s an order of magnitude better: https://thelastdriverlicenseholder.com/2019/04/21/new-waymo-... https://medium.com/waymo/waymo-iaa-frankfurt-2019-b3cca36d84...
One would imagine this early announcement comes in reaction to their recent “valuation” cut: https://www.cnbc.com/2019/09/27/waymo-valuation-cut-40percen...
If Waymo starts driving without safety drivers en-masse, even on small fixed routes, it will signal certain takeover metrics have been hit. It will be interesting to see if Waymo uses existing cars (the ML and planners just got better) or if they use new hardware and signal a certain lidar improvement is necessary for achieving a viable takeover rate.
The signal won’t be clear but should be there. And then it’s up to the rest of the industry to show how fast they can react to the change.
Haven't seen a recent study/article around cost/ROI (not sure if these have gone down significantly) -- any recommendations?
2) Check Alphabets/Googles bank balance. They have a few bucks they can invest for now.