Chess is a very complex rules game, while Checkers is quite simple. Waymo has a complex approach to self driving:
- Expensive suite of sensors
- High resolution maps of operating areas
- Remote operators standing by
While Teslas approach is simple:
- Capture a bazillion miles of camera footage, feed into AI, profit?
- Unpaid volunteers teach the AI safe driving
- Car has only a basic map for routing, the rest is inferred in real time from cameras
Waymo’s successful approach scales linearly. They have to high-res map every city they want to operate in, and they can gradually bring down the cost of the sensors. They will require fewer remote operator interactions over time.
Teslas success is more difficult, but it scales exponentially. They already produce vehicles at scale and full control over all the equipment on board. The existing fleet would be able to participate as well. If they succeed, they may want to offer buy-backs for customers who didnt buy FSD - the cars would be worth more to Tesla than the owner.
In both checkers and chess, the player gains super powers for reaching the other side of the board. Time will tell who reaches the other side of the board first. They are playing different games on the same board. Okay that’s fair.
Tesla will hit walls with rain and snow. Cameras will fail before other sensors in those conditions.
Radar and Lidar also get a lot of noise from heavy rain or snow. Fog can be just as bad. Some conditions just aren’t safe to drive in, regardless of who’s driving. I don’t think either of them are trying to design a system for those conditions.
On a personal note, I have no interest in getting a ride in a self driving car. I do have an interest in an empty car that can drive itself. Drop myself off at the airport, valet parking downtown, easier to share one car per household, river shuttling, through hike shuttling - I would use it a lot. I understand the more profitable goal is taxi services, but I don’t want that. So in my narrow use case, I hope Tesla succeeds since that approach can be used on personal vehicles anywhere while Waymo is strictly city taxis, which I don’t use.
The USS they took out of Teslas were at least a second measurement system.
Wavelengths with decent water transmitability exist.
We should abandon self driving cars and instead make the roads more safe and provide multi modal transportation options like trains.
So the article repeats, several times, “waymo relies on remote operators”. I don’t think the author knows what “self-driving” means.
I’m generally the one defending Tesla in these threads but on Waymo’s defence; the remote operators aren’t actually controlling the car. Instead when faced with a challenging situation the car “calls home” for assistance basically asking a human to take a look at the situation and help it to decide what to do. This might be a car partially blocking the lane of traffic cones placed in a weird manner so the car justs asks for assurance that it’s okay to proceed. In the most difficult situations the remote operator can suggest a route for the vehicle to take but the decision on what to do is ultimately on the vehicle itself.
“self-driving”
Someone else, forreal - driving
A couple years ago, we used to joke around the shop “Of course they used AI - An intern.”
The author is very well aware of this dilemma, in fact that topic is the center of his article, and he is making some good points about why real autonomous driving might still take a long time until achieved.
Besides that the cars are constantly getting around without a designated driver. For the technology and for the industry that is a huge breakthrough.
And he forgets to mention the precise mapping required too. He also left out the terrible experiences Waymo has had with revoked permits, cars disabled by traffic cones, and multiple traffic stopping glitches where intersections were blocked for hours.
Yeah but the waymo ceo doesn’t shitpost on Twitter so people here don’t get front page hyped up stories every single time things aren’t perfect
I’m making a prediction right now that real self driving will eventually rely on people from impoverished countries remotely operating the cars of wealthier countries. Sort of like how AI training data is combed through.
Yeah Waymo has been silently ticking away over here for years. On the east side they are all over the place. It will take longer to get to freeway speeds but I think Waymo’s approach is far safer. But fuck Google too
Both are risking the lives and safety of the non-consenting public as they beta test 2-ton vehicles on public streets. Damn them both.
They may not be perfect yet, but if their track record is safer than a human driver they aren’t any worse than any of the other assholes on the road.
Millions of human drivers are risking the lives and safety of the non-consenting public, too, but we aren’t advocating for stronger driving tests to keep bad drivers off the road. We’re just bitching about someone else trying to solve the problem because it isn’t a perfect solution on day 1.
Their track record isn’t safer than a human driver… because their system is a mechanical turk.
If their track record is better it’s because it’s a record at a much lower scale, under more controlled conditions, and kept by companies with a vested interest in it appearing good. If Boeing can hide flaws in flying vehicles carrying hundreds of people per trip I think these companies can hide flaws in cars.
we aren’t advocating for stronger driving tests
Why aren’t you? Wouldn’t that be the most logical answer?
My country has had a very bad traffic safety record, among the worst in the EU, and that was one of the things we used to improve things, along with harsher consequences.
There’s also another solution, reducing people’s need to drive. Public transportation could be improved by a fraction of the money that goes into these self-driving endeavors.
Just adding “AI” to something may look cool and even make sense at small scale but ultimately completely fail in real life. “Sometimes knives kill people, let’s put AI in knives that will retract the blade instead of cutting someone”. Does that sound plausible too?
I wonder what you then think about people who drive after heavily drinking or taking drugs. To be honest, I have more faith in technology than in humans.
Not to mention that self driving can probably solve some other problems too, like traffic jams caused by erratic driving behavior of humans, etc.
If you have vehicle to vehicle communication, it is possible to adapt the speed of all the vehicles on the street to avoid them being stuck in a traffic jam.
Driving while inebriated is illegal, self driving is not.
Traffics jams and erreactic behaviour could be fixed if everyone is in a self driving car, but at that point it woild be far more energy effecient, environmentally friendly and cheaper for society to build electrified transit instead.
If you prioritize the street so that only self driving cars are on it and they need wireless communications to function, how do other road users like cyclists and pedeatrians safely use the street?
Self driving cars are not here to make your life better, they are here to make a handful of people rich.
I tend to disagree here. For example if you have vehicle to vehicle standardized communications, vehicles can communicate between themselves the location of cyclists, some road obstacles, etc. generally making the roads safer and reducing the number of fatalities.
Yes, they will make some people more rich, but is this a legitimate reason to obstruct technological advancements? I am sure people were thinking the same way at the cusp of electrification, or automation of some factories, where machines were augmenting the human labor and in the process making those people redundant.
If we think the same way we should never abandon coal power plants and mines because miners might lose their job, right?
There are greener, more energy effecient and more socially fair ways to get the same results than selling everybody a high tech steel box.
What do those options matter if nobody is developing them and they only work in dense cities? You might as well be arguing for Star Trek-like transporter technology here.
This is literally the only way we’ll ever get self-driving cars. You have to test them in real life. Simulations and tests tracks can only take you so far. Yeah it’ll probably cost the lifes of some number of people but this will be greatly outnumbered by the amount of lives saved when the technology actually starts working as intented. It’s not like human driven vehicles are exactly safe for pedestrians either.
Also, when a self-driving vehicle fails it almost always means it ends up getting stuck somewhere or blocking the road. It’s extremely rare for it to cause an accident, though that does happen aswell.
Yeah it’ll probably cost the lifes of some number of people
Easy to say when those lives doesn’t include yours or anyone you love/care about.
So you’d rather more people die in avoidable traffic accidents because we weren’t allowed to develop this technology?
I’d rather have people avoiding using cars at all, adopting mass transit solutions instead.
Great that’s admirable, but that isn’t going to happen because it doesn’t work for most people and there is no political capital to make it happen, so what then?
How could anyone know that? It just as well might.
It’s a fallacy to think we can build a perfect world where all bad things can be avoided. With all new technology comes downsides. We’re already losing 80+ a day in the US alone because we don’t have self driving cars. It’s far more likely for someone close to me to get killed by a human driver.
I don’t think public deaths is a valid cost for creating self driving cars. We could be builidng safer and more effecient transportation systems. Some billionaire is going to make even more money because they were allowed to use the general public and city streets as a testing ground for their product. This is not fair to the family or the people who are injured or killed by self driving cars.
There are currently 80+ people dying every single day just in the US alone because we don’t have self-driving cars. Not developing that technology is just as much of a choise to let people die than going forward with it. I’d argue it’s the moral thing to do. People are awful at driving. As a fan of cars I like to go sit by the freeway watching them passing by several times a week and the number of people driving 120kph while staring at their phones is mind boggling.
Not only that but virtually all of those vehicles are going to be electric as well so that also means less people dying because of air pollution. Then there’s also the fact that it’ll bring down the cost of taxies immensely as well as allowing private individuals to let their vehicle go do ride sharing for the day instead of sitting on the parking lot of their work place unused. There’s just too many upsides to it. Also it’s not like passengers getting killed by rogue self-driving vehicles is a particularly common occurance despite the technology still being at it’s infancy. This is the worst they’re ever going to be.
The same problem could be fixed with electrified transit and walkability. Transit would also be even more environmentally friendly.
Plus we could still develop self driving cars but do a lot more testing before we set the public as the guinea pigs to see if they are safe.
Id also argue that we cannot claim this is the worst self driving will get since self driving cars are only used in a few areas right now.
Like I said; there’s only so much you can test on a closed track. At some point you must start doing that in the real world. Pedestrians getting killed by experimental self-driving vehicles is not an actual issue we’re dealing with right now but more like a theoretical possibility of what could happen in the future. There are only a couple of such incidents recorded ever. That’s not a good enough reason to not continue with it.
What I mean by them now being the worst they’ll ever be is the self-driving technology itself. It’s constantly improving and the trend is towards better. The technology we have right now is the worst it’s ever going to be.
And Cruise is playing bumper cars
More like GTA
Elon ripped out the LiDAR, I don’t have any faith at all in Tesla’s FSD.
But the self driving taxi fleet is coming anytime now /s
Elon ripped out the LiDAR
No he didn’t…He never even installed it in the first place.
I had to look this up, but you’re mostly right. They never really did use LiDAR. They did use other types of radar, which were removed or disabled. In any case, they (Elon) asserted that neither radar nor LiDAR was really necessary.
However, that was mostly a couple years ago. In the past month or two they actually have begun buying up tons of LiDAR.
Also, they were sued over FSD in court and their lawyers are now arguing that customers should’ve known that cars without LiDAR are not capable of reliable FSD.
Humans can drive just fine without lidar aswell. Road infrastructure is designed for vision. The car not being able to see is not the issue. It’s teaching the car to understand what it sees and how to deal with it. Lidar doesn’t help you solve this issue.
aswell
Never a word, my dude.
I’ll try to remember that!
Unlike human eyes Tesla’s inconveniently do not come with a supercomputer installed that is able to interpret the optical data reliably. With the compute power we have available Radar based navigation is the only one that produces reliably safe results.
Except we don’t have reliable radar navigation, what are you smoking? Are you just making things up you think should be true?
It depends on your definition of “super computer”, it used to be any computer with performance over 1 gigaflop, which today would probably include most smart phones and the built in car computer in the Tesla.
But regardless of semantics, I think your point holds, humans are a special case and computers can’t do that yet.
Bro he was talking about your brain And there is no supercomputer on earth able to even approach it, and regardless supercomputer is a rather relative term tbh
Bro he was talking about your brain
Yeah, I caught that… Obviously.
and regardless supercomputer is a rather relative term tbh
That was essentially my point. Keep up.
Atleast Model S and X with hardware 4 do have radars in them.
Huh, guess they came around in the end then.
But who’s playing Clue?
It was the car with the car on the road. The car did it with its own body
And in the news this just in… Tesla runs over checkered flag and flagman at Daytona. Shortly after, it burst into flames. As it burned it was discovered that the car’s emblems melted into the shape of Toyota emblems…
Comparing Tesla with Waymo is stupid. They are doing fundamentally different things, and people like this author don’t realize that. Waymo’s technology, like a few self-driving products from Ford or GM, rely on having a centimeter level 3D scan of the road ahead of time. This allows a crap ton of pre-processing so fewer decisions need to be made in the car. It’s a developmental shortcut, but it also means their cars will only work on roads that have been scanned and processed and approved ahead of time. Tesla’s system doesn’t pre scan roads. It makes all the decisions on the fly based solely on what the car is seeing as it drives. That means that it can theoretically work on any road, in any situation, without advance preparation.
Tesla’s approach tackles a MUCH harder problem. And that must be considered when comparing the two technologies.
Otherwise it’s like looking at two people at the gym, William lifts 25lb weights and can now lift them 10 times, Tom lifts 250 lb weights and can now lift them 9 times, and saying that William is in better shape than Tom because he can do more reps. No, Tom is in better shape because he is lifting a lot more weight. Even though he can’t lift it as many times, he’s doing a lot more work in his workout.
Removed by mod
Rapid deployment and iteration belongs on race tracks, not the public road.
I used to be so excited for self driving cars, but my naive younger self assumed they’d actually make sure they’re safe before putting them on public roads.
I was wrong.
We’re getting those American robot cars
What’s a few 100,000 deaths compared to progress?
/s
I’m starting to get the feeling that “X is playing chess while Y is playing checkers” is an indicator species for a terrible take.
Much like everytime you see “nothing burger”
It really is an insult for checkers as a game. It is a common misconception that it’s simple. The game has surprising amount of depth, and the saying “x is playing chess while y is playing checkers” should really die.
X is playing chess while Y is playing tictactoe would be a better analogy.
I usually take the chess/checkers idiom to be more like “the left hand didn’t know what the right hand was doing” Not that one is smart and one is dumb, but that they’re going in completely different directions and playing by different rules.
Hmm, maybe I’ve learned the meaning wrong. English isn’t my first language :)
Chess has roughly 10^44 positions. Checkers has roughly 10^20.
That means under that metric, chess is roughly 24 orders of magnitude more complex as checkers.
Tic tac toe has roughly 10^3 positions, or 17 orders of magnitude simpler than checkers.
In other words, the complexity gap between chess and checkers is larger than the gap between checkers and tic tac toe.
Agree. Checkers has been solved, Tic tac toe has been solved on xkcd. Chess has not been solved
Maybe they should compare playing chess with playing Go.
The number of legal board positions in Go has been calculated to be approximately 2.1×10^170, which is far greater than the number of atoms in the observable universe, which is estimated to be on the order of 10^80.
My point is that checkers actually still is very mich complex. Tictactoe is not and every board position can reasonably be managed by a human.
With checkers, that is unfeasable. That’s why I am of the opinion that checkers is unfairly treated as “the simple game” when for humans it is far from simple.
Waymo is literally being investigated for self-driving incidents lol… They’ve also only been able to operate in a small area. I wouldn’t really call that chess.
