So promising that it will be not just solved, but productized into a polished consumer gadget, in 5 or so years, was an astonishing, unbelievably foolish promise.
And worst of all, yes, it came from people who should have known better.
And as a quick aside: there's much talk about "ethical AI", but here's a more common AI ethics failure: taking money from funders, shareholders, and customers, with promises of imminent deliverables, while knowing full well that the AI advances required for that haven't happened yet, and there's no evidence for when they will, if at all.
To add insult to injury, the manner in-use today of modeling humans as essentially just another kind of dynamic object breaks down extremely quickly once humans are normalized to the presence of autonomous robots. The humans change their own behavior model to achieve their own objective function. They're not dynamic objects, they're learning objects, and now your AV models (which will include human interaction and reaction behaviors on purpose if you're it right and on accident if you're not) breakdown along an entirely new longitudinal axis.
I think the AV market fundamentally broke when GM acquired Cruise for $1B seemingly out if nowhere.
1) Because the reality is that the DARPA Urban Challenge result was not assuredly generalizable, and it may be the case that there's still basic science to be done before the domain is just better, cheaper sensors & high-performance compute away from being productionizable.
2) It's not at all clear that deep learning is really iterating toward a solution to this problem either, but the improvements to methods and hardware have produced the ability to make very compelling demos, that even the purveyors of themselves might believe in, and show iterative improvements on the previous result, thus fueling more investment into something that isn't even known to be possible.
3) But, none of that would matter if GM hadn't decided to make a power move in throwing $1B at an acquisition and in one fell swoop turned an early stage unproven scientific research market into a super, super frothy capital market where it looked like anybody at any moment could be a unicorn without any clear reason or any fundamentals.
Investors went insane and the expectations and promises followed them.
I say this while on a car ferry. Getting on this boat required me to navigate several strange road markings and obey a half-dozen hand gestures from staff, including several that were contrary to the painted lines. No AI is even contemplating car ferries.
There aren't many car ferries around where I live. But there's plenty of construction...
That's a huge difference. Maybe you can address it with some sort of remote OnStar-like system but now you're forcing a remote operator to jump into an unknown context and take actions that were too hard for the AI.
I don't think we will ever change road infrastructure in the same manner on any appreciable scale. (Hell, there's a pothole in front of my street that's been there for 5 years).
This holds if you expect self-driving cars to be able to work like humans in picking up "body-language" cues around other drivers. But we can also approach the problem the other way around and set better fixed rules for how driving needs to be done and adjust the road and car infrastructure to make the problem simpler. Human drivers get away with breaking so many rules that the problem becomes much harder than it needs to be.
Found it: https://craphound.com/spamsolutions.txt
The parent response would have these ideas checked:
(X) Requires immediate total cooperation from everybody at once
...
(X) Asshats
(X) Jurisdictional problems
...
(X) Technically illiterate politicians
...
(X) Countermeasures must work if phased in gradually
Automating being dismissive of discussion does sound like a great idea...
> Requires immediate total cooperation from everybody at once
It requires setting stricter traffic rules and having them be followed. Something that's done everywhere in the world every year. It can be done as gradually and as locally-specific as needed.
> Asshats
We already have those on the road causing accidents. Traffic rules are there to punish this exact behavior. The fact that we don't enforce a few of them makes self-driving harder. I'm not "failing to account for asshats", I'm specifically targeting them.
> Jurisdictional problems
We already have different jurisdictions where different driving requirements exist. Having self-driving cars that are only allowed to drive in country X would be nothing new.
> Technically illiterate politicians
My suggestions was specifically to make the road rules stricter and enforced. This is not a technical issue, it's the boring old issue of what the road rules should be and who should follow them.
> Countermeasures must work if phased in gradually
Enforcing road rules works fine even if done gradually. It also makes self-driving gradually simpler.
> It requires setting stricter traffic rules and having them be followed. Something that's done everywhere in the world every year.
But you mentioned two somethings: "Setting stricter traffic rules" and "having them be followed." The very first examples in the TFA were from Argo's testing i Pittsburgh:
Recently, one of the company’s cars encountered a bicyclist riding the wrong way down a busy street between other vehicles. Another Argo test car came across a street sweeper that suddenly turned a giant circle in an intersection, touching all four corners and crossing lanes of traffic that had the green light.
Setting stricter traffic rules is certainly done everywhere in the world every year. People failing to follow traffic rules is also done everywhere in the world every day, and that's the big problem. Fully autonomous driving requires the ability to make snap decisions that may have little to no precedence in your past experience. This may be a solvable problem for AI, but it's a really, really hard problem that self-driving aficionados seem to consistently underestimate.
...
ie where we want to use cars.
I don't need to go into everything that was said, you can google it. There was one thing said, however, that I think would be appropriate to touch on here. Kamen, just like you, claimed that traffic laws and the streets could be restructured to better accommodate the segway. That was the instant I knew the segway would not be taking off for a long, long time. If you need for people to restructure their laws and cities solely to accommodate your new technology, you should probably work a little bit harder on perfecting your technology. Because restructuring society's laws and transportation infrastructure simply to use your technology is probably not going to happen. The only time you get a radical restructuring of that nature is when the invention frees you from needing the infrastructure at all.
To be fair, this did happen -- with cars.
But that also sets the bar: to warrant re-architecting cities, an advance needs to be as superior to cars as cars were to horse transport.
Even restructuring a citys roadways for bicycles--a long-term, proven good technology--has been incredibly slow going. And this with something governments are actively trying to improve.
It will be decades until there is substantial enough change that self driving cars are viable as described in the great-grand-parent. And that's if it moves quickly.
https://news.ycombinator.com/item?id=18447674
What's next? Walls on sidewalks? Computers are not smart or dumb. They are machines. Anthropomorphizing them ("self") is foolish.
You're attacking a strawman. I'm not saying the solution to self-driving is to change the environment completely to make the problem trivial (i.e., walls on sidewalks). I'm saying that some of the poorly defined and even worse enforced rules of driving could be worked on and that would help self-driving as well. How much would be enough to bring it out of being AGI I'm not sure, but it would definitely help.
If you want to instead talk about making it safer for human drivers, fine, that's a different subject.
Human sensory augmentation will save lives, but that has nothing to do with the marketing ploy that cars can have self.
I haven't been in an accident in over 15 years
IMO, if allowing self driving cars on highways costs less than 100k per mile it’s a rather trivial expensive at 16 Billion in the US. Extending that to every road would be much harder to justify. Similarly, developing something like a set of more clear hand gestures for directing traffic would not be a major issue.
Honestly, the need for expensive changes seems unlikely, though some changes such as paint choices for lane markings would probably increase safety or efficiency.
"Yeah some context missing here, but it did make for a fun headline.( said the same thing about my own driving.) The point is that autonomous driving, like human driving, will always have constraints."
To move the needle - towards a true tipping point? - autonomous vehicles don't have to be everything to all people all the time. Certainly for a significant number of driver miles it's moderate distances at moderate speeds.
Finally. Something as simple as a vehicle being able to deliver itself (from some central hub) to your door and you drive it (non-autonomously) from there would be significant. That changes the ownership model. It changes parking requirements, etc.
Yeah, the holy grail might be a ways off. But there's plenty of (pardon me) disruption between now and level five.
That said, I don't think the most advanced AI is even close to a 2-year-old.
If the kid appears in the road from behind the truck, the computer can handle slamming on the brakes very easily - probably faster than a human can - without understanding the kid any better than a group of LIDAR points or a rectangle of pixels labelled 'obstacle' by a neural network.
But if you want to brake before the kid appears from behind the truck? For that, a fully attentive human driver will be making a bunch of estimates about what the kid is doing, whether they seemed to have noticed the car, how old they were, and so on. In other words, applying a theory of mind.
Needless to say, if the latter is a must-have feature, that's a pretty hairy problem.
Of course, it's possible the decrease in deaths from being fast on the brakes in simple situations will outweigh the increase in deaths from lacking a theory of mind in complex situations. If that's the case, the self-driving car programmer's job would be a good deal simpler!
Children playing in the street are a common occurrence in residential areas, I see no reason why you would not develop a set of rules and heuristics to handle them. Identifying a pedestrian as child, and knowing whether it is running or playing, is well within the capabilities of modern computer vision.
So ... a theory of mind.
Is this type of crowdsourced driving data a crucial part of achieving Level 4+ self-driving? If so, it seems that driving around the same six city blocks of SF or cruising down Central Expressway in MV is going to produce diminishing returns in terms of producing measurable progress.
I don't think such a system would catch a false negative like the above, where the human would slow down cautiously but the self-driving system would do nothing. That situation is indistinguishable from a human slowing down to read house numbers.
To realize the problem, the system would need a full model of "what would the car be doing if not for the human input" in order to find a later point of alarming divergence.
this is bad, but not quite as bad as it sounds. most of the time, you only need to stop as fast as the car in front of you. it's pretty uncommon to encounter a stationary object in the middle of the travel lane. in fact, outside of driving on surface streets in the city, I can't remember the last time I had to avoid a stationary object in my lane.
I'm much more worried by how closely people follow the car in front of them, regardless of visibility. many leave barely enough room to react at all.
While this usually will be fine, there are definitely issues when something stationary does pop up.
While caravan-ing to Yellowstone with 3 vehicles, all traveling in the center lane of the freeway, we encountered a small car with a passed out passenger in the middle lane. My bro-in-law swerved with some room to spare, immediately behind him I swerved with basically zero room to spare, and my father behind me (luckily for them in a Suburban, but bad for the man in the VW) had no chance- I was immediately looking in my back mirror knowing what I was going to see.
There is a decent likelihood that a machine could have swerved in time, but nearly zero that a typical human could/would have in our typical following patterns.
Humans route around inefficient practices. Just as human driving speeds are typically unaffected by posted traffic speeds, they will optimize for their typical experience over written codes for how they drive.
One example is that there are a number of small towns with two way streets that are parked on but only the width of two cars. That creates bottlenecks where cars can only travel in one direction at a time. This might seem like a disaster and it certainly wouldn't work on a busy city street, but in these locations everyone adjusts and when two cars approach a choke point from opposite directions drivers are really good about being cautious and pulling to the side to allow the other party to pass.
There are probably thousands of these local quirks around the world. Handling all of these situations effectively in a fully self-driving car will take an advanced AGI.
A object would need to be substantial enough to cause an accident and then roll into or fall onto a highway. That’s far from a 1 per million miles of driving situation. Remember, something falling off a truck would also take a while to slow down.
But human drivers hit deer all the time, and it's often unavoidable. An autonomous driver is probably more likely to miss a deer than a human, due to substantially better reflex time.
To me, every brown mailbox with a white reflector could be a deer coming to the road, to infrared, with a larger lens, it should be able to tell a lot better.
Autonomous driving doesn’t need to be perfect because we don’t need it. With sufficient alternative systems the only reason for driving will be for hobby (and we don’t want that automated anyway) and heavy load work (like agriculture, mining, or logging) which is already heavily automated.
I'm not walking halfway across the country just to visit my mother.
The sum of these alternatives will almost always outweigh driving in terms of benefits with a notable exception of convenience. So if you are willing to sacrifice convenience when you want to visit your mother, you will almost certainly get there faster and more economically (with the right systems in place) then driving.
With that said, yes there are faster ways (albeit still less convenient) of getting you outside of said city. You might have to change your mode of transportation a couple of time (I said it was less convenient) but it will still be faster with the right systems and infrastructure in place.
1: This is not an unfair scenario because this is already almost the case for all of the non-alternatives.
Kiss an extra two hours of your day goodbye just to get to and from work (assuming you have a job). Going to the doctor is a nightmare (assuming you have medical insurance). It sucks a lot, unless you're living in a city so dense that cars are impractical.
I hope your local politicians agree with me that expenditures going into making these alternatives are money better spent then waiting for the technology to dedicate highway lanes for autonomous vehicles.
That is I hope they agree that your situation of not being able to go to the doctor within a reasonable amount of time takes precedence over people wanting to sleep during their 8 hour highway trip but are unwilling to take the bus for some reason.
Yes actually, although there are some philosophical objections to computer assisted proofs: https://en.wikipedia.org/wiki/Four_color_theorem#Proof_by_co...
While “ai mathematical proof” gave me “Google AI system proves over 1200 mathematical theorems”: https://mathscholar.org/2019/04/google-ai-system-proves-over...
So… yes, that’s fine?
It's always just around the corner, and has been since the dawn of AI. Maybe we will get there some day (I never say never), but cautious realism has never been the AI field's strong suit.
Radio Yerevan was asked: We are told that communism is already seen at the horizon. Then, what is a horizon?
Answer: A horizon is an imaginary line which moves away each time you approach it.
A lot of people are working to create a car smart enough to drive humans anywhere. How many people are working to create a car smart enough to refuse to drive because it's feeling inspired to write music instead?
They do NOT need to be perfect. They need to be BETTER than humans. That means less accidents, and by extension, less deaths and injuries as a result of automobile errors.
Forget the snark: you don't say? :o
Some of that logic is literally milllenium old unsolved logical dilemmas ("I have 2 choices and both involve killing people...").
I thought you referencing the media here. Build it up, tear it down. They quoted and amplified so many people who suggested it was possible only to print articles later that point out the it just ain't possible.
That's equivalent to saying AGI is impossible.
It's a pretty big statement.
Super AI's I think are achievable and will be developed. I just don't know which will come first level 5 or cold fusion that's a toss up. If everything though has sensors including bikes, pedestrians, cross-walks, roads, etc that can detect what is on the road, send out a warning signal to cars travelling towards that intersection then that also takes out some of the problems - but that requires a very smart grid.
On a long enough timeline though there's probably not much humankind can create. Though it could be 10 years could be 50.
“Never” is kind of a strong word in this context... My personal belief is that we could very well have level five autonomy within the decade, using only two video cameras and two microphones behind a windscreen. But this would of course require major progress in A.I., of the sort that may also not happen in a hundred years.
Getting to level 5 will require "V2X" technology - vehicles communicating with other vehicles and the road/traffic infrastructure itself.
We will need a forward-thinking government to start building sensors and communication tech into roads, stop lights, parking spots, and a secure/interoperable internet for all these sensors to communicate autonomously.
Currently the cost of implementing such a system is way higher than the return to drivers who can...what, work on more powerpoints or facetime while they're in the car instead of driving? Not to mention the huge security and safety and liability risks this would open up.
That's why we solved this problem with public transportation in the past - hire one person to be the "V2X sensor" AKA bus driver.
>Waymo’s CEO, John Krafcik, has admitted that a self-driving car that can drive in any condition, on any road, without ever needing a human to take control—usually called a “level five” autonomous vehicle—will basically never exist. At the Wall Street Journal’s D.Live conference, Krafcik said that “autonomy will always have constraints.” It will take decades for self-driving cars to become common on roads. Even then, they will not be able to drive at certain times of the year or in all weather conditions. In short, sensors on autonomous vehicles don’t work well in snow or rain—and that may never change.
https://www.cnet.com/news/alphabet-google-waymo-ceo-john-kra...
https://www.spectator.co.uk/2018/07/the-dream-of-driverless-...