Tesla AI Day Event Did a Great Job Convincing Me They're Wasting Everybodys Time | Hacker News Reader

Tesla AI Day Event Did a Great Job Convincing Me They're Wasting Everybodys Time

jalopnik.com

56 pointsjocker1249 comments

> Here’s the root of the issue, and it’s not a technical problem. It’s the fundamental flaw of all these Level 2 driver-assist, full-attention required systems: what problem are they actually solving?

This article was so good and hits the nail on the head in so many spots. I actually find I can zone out and think about other things BETTER without autopilot on. My brain drives autonomously much better without me having to play the wheel force game and constantly be in anxiety at every AI controlled turn.

My brain has had a lot of training sets over many decades and drives autonomously quite well.

Hmm… not me. Keeping myself centered in the lane is tiresome (first-world problem, I know). My car only requires that my hand be resting on the wheel. No games.

It’s a bit annoying to be perfectly centered at all times, I will admit. You don’t notice until you’re no longer doing it, but lane positioning is very dependent on what’s next to you. For example, being centered in your lane next to a big rig actually feels uncomfortably close to it. Luckily my car lets me adjust where I am in the lane without disengaging, but I will turn off lane keeping entirely if I find I’m fighting it too often.

Now, when someone makes a system that does smart lane positioning, I’ll be set!

I still have zero interest in the city stuff mentioned in the article.

I suspect the difference comes at least partially from the amount of electronic in between the stearing wheel and the road, preventing most feedback, including unconscious.

There is a direct mechanical link between the steering wheel and the car's front wheels. It's not steer-by-wire.

There have been many variations of assisted steering between mechanical assistance and steer by white, though.

> Keeping myself centered in the lane is tiresome (first-world problem, I know).

Only because first-world nations are a bit better at enforcing lane laws. India, Iran, Indonesia, etc... the lanes might as well not exist.

I recognize this isn't the point you were trying to make, but it was what my mind was immediately drawn to lol

In India you create your own lane.

That's why traffic and traffic jams are so bad there. What a waste of human existence to be stuck in things that are solvable with adhering to some lines of paint.

There are quite a few models in the market that are not Tesla and have lane centering assist.

Something doesn't add up, and I think it's the author not being able to perceive the problem here.

For example, OpenPilot seems to be very comfortable with Level 2 driver assist, and so do their customers - and their position seems to be make driving "chill", like you're commuting more relaxed because the car is doing the boring stuff, yet you need to be present to make some decisions.

So clearly this is solving a problem for a lot of people.

What people need to understand is that if you don't understand these products, then maybe they aren't meant for you - and that's OK!

They are not trying to please everyone, yet that doesn't mean there's not a problem being solved for some.

>the car is doing the boring stuff, yet you need to be present to make some decisions.

That sounds like a useful thing, but it doesn't sound like what the article was about, and it doesn't sound like a thing that exists.

I think a way to rephrase the problem is that humans are some things and machines are good at others, and one of the things that machines are good at is instant safety cutoffs.

The self-driving attempts I've heard about rely on humans to be the safety cutoff.

Cars are often milliseconds from disaster, because of the close lateral proximity of objects while driving. Like concrete Jersey barriers.

You can create space in front of you, as they tell you in defensive driving class, but you can't create 360 degree space.

This is the first I've heard about openpilot, but the first thing I read about it is that comma.ai released a device called "comma one" in 2016, a month later the NHTSA asked for information on it, and they said "lol, never mind, it's cancelled".

I guess that it's open source now to avoid liability.

You should look more into comma.ai, they actually just released "Comma Three" few days ago, the first version with custom hardware and not only off the shelf parts. They actually track the driver and warn you to take control (when you're not looking at the road for example).

They seem to be navigating well in terms of liability, namely saying they want to remain Level 2 for the foreseeable future. It's like an upgraded lane assistant/cruise control (of course it seems way better, not trying to sell it short).

From what I've been reading, there's a lot of people that would much prefer this option over giving themselves fully to self driving, or even the current solution that Tesla provides.

Btw I'm no expert on the comma project, I don't own any, I'm pretty sure there's plenty of people here that know way more then I do and probably own comma hardware. I just keep an eye on what George Hotz is up to.

I have an L2 car and absolutely freaking love it.

I still need to "drive" but I don't need to actively provide input every single moment that I'm driving. Basically, indeed to mimic the cars movements rather than force the car where I want it.

It significantly reduces fatigue and generally increases my enjoyment t of driving.

> That segment of video was kind of maddening to watch because that’s an entirely mundane, unchallenging drive for any remotely decent, sober driver.

I vehemently disagree with this. Notice how at 38:24 how it slows down for the pedestrians, most importantly the child, even though it had clear right-of-way? Why? Because children do stupid shit. That's why school speed zones exist, and why they had to double fines in school zones: because these "remotely decent" human beings are incapable of correctly assessing their skill, and regulating their behavior. People are concerned about how AI determines who to kill in a lose-lose situation, but humans routinely send the trolley towards other humans for the sake of their own time.

How many people do you know who actually adhere to the speed limit, who actually maintain safe following distance, and who actually practice defensive driving?

Barely competent and trivial situations are already far beyond human capability.

The best comment:

"Rocket science IS easier than self driving.

1) There is usually only 1 rocket in the air at any moment

2) Non-rocket vehicles are easily diverted well out of its path

3) A 3-4ft deviation in position doesn’t mow down people

4) Only some of the vehicle has to return in one piece

5) The final and most important fact: The weather is basically perfect for every launch."

I really don't understand the antipathy towards systems like this. Even at their worst they are 5 to 10 times safer then human drivers. That's the point. That's the only point that should matter. I understand people want some robocar utopia but it doesn't all happen at once. I will happily take a 10x reduction in driving deaths.

0.9x as safe as human drivers.

https://www.forbes.com/sites/bradtempleton/2020/10/28/new-te...

And as the system gets better, that number might go down instead of up since we won't be able to maintain focus.

He’s making a useful point, but doing a really poor job of it. That article creates a graph from thin air as far as I can see based on arbitrary assumptions via cherry picked estimates to make a point. For example only listing city driving and limited access highways.

“If we reduce the estimate ratio of accidents/mile on freeway and non-free to just 2.3:1 (as a number from another source suggests) the chart shows Autopilot slightly below non-Autopilot until Q1 2020, then being around 15-20% better in 2020”

So, he’s also estimating it may have been slightly worse before 2020 and significantly better in 2020, yet he’s happy to post the 0.9 number based on what exactly?

Somewhere there should be stats on taxi and bus driver accidents.

I think that is what self-driving should be compared to, not the average person.

It's "applying for a job" as a professional driver, not as a simulated drunk, texting, shaving, eating, feet-on-dashboard average American driver.

Not too high a bar.

I agree it’s a good idea to have higher standards for self driving Taxis. However, delaying self driving cars that would be safer than the average driver would cost lives. Mostly I just want drunk drivers taken off the road ASAP.

Suppose we have a lot of self-driving vehicles that are slightly better than drunks, and the worst drunks use them. Maybe they can be hooked up to breathalyzers or something.

How can we ensure that the increase in the amount of driving (riding) that drunks do, doesn't more than cancel out the reduction in the rate of accidents?

What if, the self-driving being available to non-drunks, they use it because it's fun and futuristic, and that increases the amount of accidents even more?

> slightly better than drunks

We are well past that point. At even just 0.08 BAC a driver is 7 times as likely to cause a fatal accident and it gets dramatically worse as BAC increases. Self driving are close to the norm right now, even if they where say 20% more dangerous than an average alert driver it would take a massive increase in miles driven by drunk people for things to end up worse.

Anyway, the penalties are currently the same with and without self driving, hell sleeping in your car while drunk is illegal. As long as the laws stay the same things should stay about the same while saving 10’s of thousands of lives.

A drunk driver might be 7 times, or 70 times, or 700 times as likely to have an accident. That range is plausible to me.

But estimates of the proportion of trips under the influence of alcohol that I was able to find were around 0.03%. (111 million out of 411 billion)

So I don't see a logical reason to rule out their driving increasing by ten times, or 100 times, or more.

And I don't see what the justification would be for assuming a self-driving system is within 20% of an arbitrary standard. Because drunk drivers are so bad, as you point out, there's a lot of space between "as bad as a drunk" and "as good as an average non drunk".

I think it's completely plausible that a system could be ten times or a hundred times worse than an average human and it not be obvious. That's because an average human has one fatal accident in approximately 100 lifetimes (in the US). Or, for another reference point, drivers had fatalities about 7 times more often per mile, in 1950 than now.

Without saying I know exactly what numbers to use, I'm pretty certain that betting on things matching to within 20% is unjustified without, well, a specific logical justification.

However, regarding penalties being the same, this seems like a lack of incentives that would mean neither lives saved nor lost.

Also, I don't think saving tens of thousands of lives per year is a possibility if we're talking about alcohol in the US. I believe there are only on the order of 10,000 fatalities per year.

I note that millions of people (again, in the US) seem to think 10,000 covid-19 fatalities in only a few weeks is not worth even wearing a mask to stop.

Some of the things I read while writing this:

https://en.wikipedia.org/wiki/Motor_vehicle_fatality_rate_in...

https://www.cdc.gov/transportationsafety/impaired_driving/im...

https://www.worldometers.info/coronavirus/country/us/

> around 0.03%

People don’t drink that much to really push the numbers all that high. 100x as much would mean people spent 3% of there time drunk while making trips. That would take drastic changes in behavior as most people generally either don’t drink or drink at home. Considering the average drunk has a much higher than .08 BAC I just don’t see this as a significant possibility.

Also, 20% was used as twice as bad as the pessimistic article about Tesla’s Autopilot posted in this comment thread.

The thing is if self driving systems where drastically worse we would have significant real world evidence of it. Basically the better the system the more data you need to collect. If they killed someone every 1,000,000 miles well there would be piles of bodies from these systems. Arguably only Tesla’s system has gotten anywhere near enough data to start doing some comparison and the error bars are still quite high. I have no problem saying near human performance on average but specific numbers depend on various assumptions. Some estimates say better, others worse that’s close enough to say it’s much better than a drunk driving.

Take Google’s completely autonomous system, it’s only at 20 million miles. So sure it’s got zero fatalities but a group of humans could easily clock 10 times that without any families.

>If they killed someone every 1,000,000 miles well there would be piles of bodies from these systems.

If they killed someone every 10,000,000 miles, they would still be "drastically worse" than the average driver - virtually an order of magnitude.

But these are made up numbers. What data is there to work with? It seems like disengagement numbers from testing are on the order of 10-20K miles.

This article says that "everyone hates" these statistics and people claim they are easy to game:

https://www.theverge.com/2020/2/26/21142685/california-dmv-s...

But that doesn't matter, in my opinion, for our discussion, because even if they are gamed to the hilt, they provide a relevant upper bound of what is possible for self-driving in the best possible environments.

I don't know if you can see it my way, but I think switching to a safety driver, aka disengagements, are what we need to be comparing to fatal accidents, if the context is transporting people incapable of monitoring, let alone driving.

I am willing to estimate the worst habitual drunk drivers may well cause a fatal accident roughly every 100K miles. That's really bad!

But I seriously think that it would be foolhardy to plan on self-driving vehicles helping, when they punt about ten times as often as the worst drunks.

Sure, there is some uncertainty, not every disengagement has to be a fatality, but as a matter of public policy, nobody should be betting thousands of lives on that just because they would like it to be true.

Comparing fatalities of a self driving system combined with a safety driver, with fatalities of drunk drivers alone, is just not valid.

Unless you evaluate the drunks with a safety driver at dual controls as well.

Something I can't judge, and you will probably have an opinion on, is how much "low hanging fruit" remains with current approaches to self-driving. If we were to agree that it needs to be ten or a hundred times better to be better than the worst drunks, can that be done without a radical revolution in the technology, through incremental refinement? I'm skeptical, I think we probably are at least past the proverbial easy 80%, but don't have specific reasoning for that.

> It seems like disengagement numbers from testing are on the order of 10-20K miles.

That’s no longer the case. Waymo is over 30k miles per disengagement, however it’s also a meaningless number.

The point of disengagement’s isn’t a proxy for accidents it’s to avoid cars simply shutting down in traffic and make it really clear the car is confused. Waymo is operating driverless taxis with as I said over 20 million miles often with an empty car and they’re using a call center to handle edge cases. People remotely pilot the car at very low speeds when needed, but the car must come to a safe stop on it’s own.

Having someone handle an edge case every 30k miles hours for is cheap enough to be bundled into a new car’s price. Worse case they pull off the road and call a tow truck, but more normally they can get the car to a place or can safely drive. Put a drunk driver in a car with that behavior and you don’t get accidents every 30k miles.

>it’s also a meaningless number

Saying something is meaningless doesn't make it so. Often things don't mean what people would like them to mean, but every observation means something.

If Waymo is operating in particularly easy environments, then it's plausible that it's not as good as a 30K mile interval would imply.

But that number definitely implies an upper bound, because if they could go longer, they would.

Do you have a source for there being no safety driver physically in the car and the car always finding a safe state on its own?

It’s the Wamo one service in a very limited area of Phenix that’s operating without a safety driver. They started in 2020 with 10% of the rides without a safety driver and it’s been ramping up.

There’s been a fair amount of coverage for example:

https://www.wbur.org/hereandnow/2021/01/04/waymo-driverless-...

The safe state thing is inferred based on the list of crashes so far. Clearly it’s not going to always work perfectly to the end of time, but it’s good enough that their testing it without a safety driver in the car which says quite a bit.

I would be extreamly sceptic of marketing statistics.

>>Even at their worst they are 5 to 10 times safer then human drivers.

Except that it's a statistic that's abused to hell and then some. For instance, frequently companies use total number of deaths per mile driven, without any consideration of what kind of miles they were, and you know, that's kind of important. Like, driving at night in certain places you are far more likely to be killed by a drunk driver, but that risk is simply extremely low when doing a low speed heavy traffic commute in the city. If your entire commute is being stuck in 10mph max traffic, how exactly can anyone argue that your driving risks include being hit by a drunk driver at 2am or a 120mph motorway crash? It's like using the statistics for "animal caused deaths" which include venomous snakes and stingrays to estimate the risk of playing with a hamster. Yes, technically it's also an animal but you're using the wrong stat for your argument.

And Tesla has in the past lumped together all vehicle deaths for marketing purposes - so, motorcycle deaths. Hopefully I don't need to explain how that one is misleading and possibly fraudulent.

Like the article said - in these kinds of scenarios that Tesla is showing everyone, you are at an extremely low risk. Or at least, if the system is still asking you to keep full attention, then really, you've gained absolutely nothing. A modern car with front based collision avoidance radar(so all cars sold for a few years now) will avoid any collision as well as a Tesla would, without using teraflops of computing power to navigate extremely mundane driving situations.

Again, Tesla tech might be saving lives elsewhere, I'm prepared to accept that argument. But in the scenarios they are showing? Nope, don't believe that in the slightest.

Wasting? More like using them to forever-beta-test the self driving system. The thing is , it's not very likely that this will ever go out of beta, might go back to alpha. If any other company was doing this, users would be demanding to be paid for this mechanical turk work. But because it's Tesla and elon, all's good.

I was a skeptic, but I've recently become convinced Elon Musk is doing an important public service, despite the hype. He's moving "what's possible" in the public mind forward. He's making something the public thinks is way in the future seem like it is about to happen now. He sold the public on electric cars. He has everyone thinking about self-driving. He's convinced the public that civilian space programs can be better than govt space programs. He has made the general public think about the potential dangers of AI. And now he thinks the public is ready for personal robots by introducing the concept and getting people to think about it. He can't deliver on all this, but he's getting people ready.

I feel like the Jetsons did 4/5 of these 50 years ago, lol.

Thats like someone inventing warp drive and saying "star trek did this 100 years ago". There's a difference between science fiction and actually making something that works.

Did you even read the original comment?

>He can't deliver on all this, but he's getting people ready.

He's not going to build this, making his contribution effectively the same as the long generation of media that has already:

>[Moved] "what's possible" in the public mind forward.

Please read the comment I'm replying to. It's not about building, since I (and the OP) doubt he'll successfully build any of these, but apparently about the influential 'getting people ready.' Which, again, he's had little more influence in 'getting people ready' than the decades of multimedia preceding him.

I've been hearing about flying cars my entire life. That's why I added the sentence

> He's making something the public thinks is way in the future seem like it is about to happen now.

People are taking it seriously because they think Musk is about to make it happen

Everyone complains it's not good enough yet and I agree. But can we look beyond the decision to put this in production and look 2 years ahead?

Why would you need vehicle to vehicle communication? They totally are on the right track to do exactly whet human drivers would do (no end to end ML yet though).

How far away I can't say, that anything is on the road as is I think is wrong, but the demo here was really really impressive and a big leap.

I wonder if they are too proud to build a "don't know that area well"-mode where it drives like grandma and is extra cautious.

I don't get it either. This is the second 'Tesla doesn't have any idea what they are doing' article I have seen on HN today. Since when are we judging technology by its first iteration or even by the concept? Such a pessimistic sentiment

I think we're judging them against what they claimed they would have delivered several years ago by now aren't we?

This is the problem, a little bit of driver assist is great, Tesla are doing a great job. So are Volvo. The difference is that Volvo aren't making grandiose claims.

This. There are thousands of people with Teslas they bought a couple years ago who were sold on a promise of fully autonomous driving. They were told that in the not too distant future, they could sleep at home while their car was out at its job as a part of the autonomous taxi fleet, driving other people where they needed to go by itself, earning them a living!

I’m not sure how many people bought the hyperbole, but the lies and misrepresentation was criminal.

Did anybody really, honestly want this, seriously?!

Did anybody want to pimp their car out for night rides to strangers who dirty the interior and wear out the car?

Did anyboby consider that if your car is driving 75% of its duty cycle that parts like battery packs, tires, suspension, bearings wear out faster?

The idea of robo taxis was always a pipe dream. Cars are not software. The more they are used the faster they depreciate.

People love making capital investments that produce passive income, yes. Let the car drive itself to the tire shop if the tires wear out.

The only part of this that doesn't make sense is that you'd ever get in your car again. If the car took itself to the mechanic and to the detailer's shop every day where someone else cleaned it, and you just cashed checks, why wouldn't you do that?

>the only part.

No part of the idea makes sense. There are currently zero examples of adaptive self-maintaining consumer accessible robots navigating the real world autonomously.

The idea that taxi rides would cover the enormous R&D costs (likely in the trillions of dollars) that would be required to turn this pipe dream into a reality is also ludicrous.

You're addressing something different - whether the tech would ever exist. The conversation assumed it did and talked about motivations.

I think a lot of people shelled out $8,000+ for FSD under the promise that full robo-taxis were coming and they could retire their cars to the robotaxi fleet when they wanted to upgrade.

"This guy is a liar and no one took him seriously" isn't a great defence, is it.

I recently rented a SUV that regulated the speed based on following distance while in cruise control (maintaining a safe distance automatically). Loved. It.

Automobiles need to be flush with tech that first prevents accidents and makes highway driving not suck...especially in relation to pedestrians and non-automobiles.

Assistive tech has the ability to make air bags unnecessary...a safety feature the government thought enough of to mandate. We need to mandate the simple tech that saves lives.

There are a lot of below average and impaired drivers.

>I recently rented a SUV that regulated the speed based on following distance while in cruise control (maintaining a safe distance automatically). Loved. It.

Same here -- a 2019 Ford Explorer LE that I drove across the country. I recently learned that its lane assist and following distance cruise control is effectively the same as Tesla's 'Self Driving.' Honestly, it's not just disingenuous to call Teslas self-driving, it's irresponsible and dangerous given my experience with similar systems.

Don't get me wrong -- it's great!! The emergency collision braking system saved me from a major car crash with a drunk driver in Vegas. But the fact that there are people fully relying on it to be futuristic, fully-featured autonomous driving _is going to kill people._ Hell, if I had turned on assist in some of the mountain roads I drove on, I might have wound up dead myself -- and drivers must be educated about this.

Reply on news.ycombinator.com