HN Theater @HNTheaterMonth

Hundreds of millions spent and nothing commercialized. Comma has spent $10M and can make your car autonomous today. https://youtu.be/Nnh5TQ60hek

⬐ bagacrap

comma's system is an automatic lane keeper. As in that thing CMU made in the early 90s. What's the relevance to autonomous vehicles?

⬐ NoblePublius

It’s much more than that https://youtu.be/mgAbfr42oI8

⬐ nojvek

Comma is great but drains the battery when when the car is stopped. I have to jump start my car almost once or twice a month. They really need to fix that issue.

⬐ cspags

Is this a widespread issue or only certain car models? I’ve been very interested in purchasing a Comma. Thanks!

⬐ nojvek

When you connect Comma with the OBD port for power, that’s an issue.

⬐ kevingadd

Nice to see Hotz own up to his own hubris here and share what seems like a pretty level-headed opinion on the space after trying to put his money where his mouth was.

EDIT: Having watched further in the video I think the best point he makes here is that the cost part of it simply doesn't make sense - the cars are still super expensive and good level 4 autonomy relies on things like carefully generated, high-precision map data that can easily become inaccurate and also are very expensive to produce.

⬐ darawk

That's a terrible point, though. The way the costs come down is by slowly chipping away at them, and scaling up production of all the important components. This is how all new technologies arrive.

⬐ kevingadd

That's true for things like LIDAR scanners, but if in the end you need a bunch of cars (human-operated for a while, maybe forever?) constantly mapping and re-mapping roads, you can only optimize the cost of that so much. If you need a bunch of humans checking and massaging data, human wages only go so low.

It's important to note that in this video he's not calling it impossible, just calling it a scam. Whether you agree with that is up to you - I think he's overstating it - but it is definitely the case that we've been made promises over the last 5+ years by many people (including him) that didn't hold up and the self-driving cars we have now are less autonomous than promised and more expensive than we were told to expect.

⬐ darawk

Right, but there are potentially more efficient ways to go about that. Particularly if you have the buy-in of local governments to instrument their roads for you.

⬐ kevingadd

Their current method is pretty efficient - getting the general public to annotate photos to mark stoplights, etc. So they've solved the cost problem by making the entire planet pay the cost for them.

I think a company like Google can do this without expecting us to directly pay the price for it. And a truly functional, high level automated driving system with high safety WOULD be highly valuable, so maybe the price is fine? But it's hard to deny that if you look at all the pieces, as a whole, it is a very, very expensive proposition.

⬐ Retric

Compare the costs of physically building and maintaining roads vs having a sensor equipped vechile drive past them. That’s orders of magnitude cheaper and thus no real impediment. Assuming it’s both needed and people are willing to pay for self driving cars.

4.1 million miles of road in the US vs 263.6 million registered vehicles. Assuming it costs 10$ / mile you could make every road in the US every day for about 60$ / year per car. And that’s with insane cost estimates anything reasonable is going to end up being cheap.

⬐ zouhair

Can you imagine the guy that invented the locomotive that told investors that it is the future you only need to build specific rails for them to work and you have to build miles and miles of them and no one can use those rails as they are meant only for trains and you also need to maintain them.

⬐ kevingadd

We already have trains and roads, if you're proposing another massive infrastructure investment where the consequences of underinvestment include 'people get hit by self-driving cars due to outdated map data', the bar you have to hit there is a lot higher than 'here's my idea for this locomotive, we build some rails in empty grassland and it'll take you somewhere 10 times faster than before'.

It's not comparable. Self-driving is about convenience and PERHAPS scale, it's not presently enabling entirely new scenarios. And if the car is super expensive (as he points out, and as it currently is), it leads to a situation where the self-driving cars are owned by central powers instead of individual civilians, so now the cost equation for driving is different and you're giving up control over your own vehicle. They were talking about releasing waymo cars with no steering wheel!

⬐ zouhair

Not really a counter-argument. The real counter-argument would be trains are not for private use, but cars are.

⬐ gmueckl

Rails and train cars were relatively old technology. By the time the steam locomotive was invented, there were already horse drawn trains [1]. So your example is not quite backed by history.

[1] https://en.m.wikipedia.org/wiki/List_of_horse-drawn_railways

⬐ zouhair

Even now we have a lot of maps, roads, electronics and much more already made.

⬐ tim333

That was quite an entertaining video. He mostly argues the level 4/5 driver free taxi type stuff is a scam because it won't work in the near term. He's a launching an open source android phone based driver assist system that tracks the drivers eyes to check they are paying attention to the road also.

⬐ tanilama

First statement is not a secret to many people who remotely works in the so called AI industry.

Second one is an opinion.

Overall, Hotz is a celebrity, he says stuff, some of them are right some of them doesn't matter. As regards to this interview, not a lot information out there really.

⬐ superkuh

He's not wrong. Self driving cars will (at least for the foreseeable next 20 years) be stuck to operating in simple non-changing environments like the arid southwest. They simply cannot work anywhere there is snow.

⬐ solarkraft

I like that he makes an "i use arch btw" joke during the interview.

⬐ ackbar03

I always thought hotz was cooler as an actual hacker in the old sense. He's got that rogue spirit

⬐ tim333

Yeah the end statement:

Zook screws and lame-o start to really falter things are gonna look good for us and then you want to buy. I have a price. I'll sell out 1 billion dollars. For a billion a lot of things are negotiable. I could do a lot of stuff at the billion that's a good endowment for our religion.

⬐ rvz

That’s the big age old irony of this orange site. Some say Musk is cool and others say ’Hotz was cooler’ but both say outlandish claims and yet still create something real which is better than vapourware. However, for drive assistance software today? Tesla and Commai are both still considered as unsafe.

Realistically, the industry is still immature and now at the assisted-driving level and unfortunately some users still conflate that with ‘self-driving’ due to the industry’s own hype of achieving level 4 self-driving capabilities. Tesla and Comma.ai directly tell the user to keep their eyes on the road, whereas the others do not, yet Tesla drivers keep getting killed. The same is possible for Comma.ai and I would consider both software as unsafe for now, but those who don’t instruct the user to keep their eyes on the wheel as very unsafe.

But they are at least aware of this. Some users however...

⬐ craftoman

Just because you failed completing a project doesn't mean everyone else's doing it wrong. I know sometimes it's really frustrating but that's life.

⬐ earthshot

I’m old enough to remember when geohot was loudly claiming that he’d cracked full self driving but was shutting it down without selling the assets because the government is mean.

It was hilarious to see most of HN treating that obvious nonsense as credible. This is hilarious too, because of the context around his previous claims.

⬐ tim333

He jokes around but the current product is largely what he promised the first time with the some of the safety flaws fixed - it doesn't replace the mirror, it tracks the drivers eyes to check they are looking unlike say Tesla.

⬐ kick

He didn't shut it down, he released everything except the dataset as free software.

It was bullshit to get out of regulations, a classic Valleyism, not "shutting it down."

⬐ BigBubbleButt

> It was hilarious to see most of HN treating that obvious nonsense as credible.

I remember it very differently. Maybe there was more than one post on HN, but it seemed like everybody was eviscerating him and basically calling him a fraud.

https://news.ycombinator.com/item?id=12840368

In fact, going through those comments now (I've only skimmed) I can't find anyone actually defending the guy.

⬐ IshKebab

Yeah but check when he first started: https://news.ycombinator.com/item?id=12492566

> Surprised with all the skepticism here. It's not perfect but this guy is a hardware, software, and AI genius and this could very well speed up the self driving race by several years or even more

> Jailbreaking cars now? Good for him! Can't wait to see this in action.

> Go Go GeoHotz!

> Pretty exciting for sure.

> If it really works like they say and can be added to any car, that's pretty huge. I'd definitely buy one. I wonder what it takes to install it ...

There are lots of skeptical comments too - maybe 50-50. I think it was more positive on Reddit though.

⬐ BigBubbleButt

There's a difference between when geohot made an announcement that he's working on a hard problem like you referenced, and when he said he'd solved the problem (which would trivially make him a billionaire) but just couldn't be bothered to deal with regulation like OP mentioned.

⬐ mothsonasloth

George makes the economic points but like most hackers/devs fails to see the humanist point (something I am learning to appreciate).

Forget about the big money, Elon Musks and the "world-changing" tech. Behind it all is a dangerous precedent to automate away our daily lifes, byte by byte.

10 years ago I had instant messenger, basic satnav, email, holiday bookings, price comparisons and online shopping.

Nowadays I have dating algorithms, live satnav, home automation, voice assistants, smart motorways, groceries and weekly sundries handled by amazon prime, cars collecting telemetry on my driving habits.

These things are influencing or completely removing our brains need to make decisions.

Through this, comes an easier, safer or happier life, however you have to think, at what cost?

Do we become dependant on machines and evolve towards this "singularity" that we hear mentioned around here often. Or do we become willing hostages to our technology?

⬐ mattlutze

Cars driving themselves means less people die on the roads. The self-driving features that are in existing fleets of cars are demonstrating this.

That's the humanist point that is being missed, and running "this is all a scam" videos when Hotz isn't even saying that seems disingenuous.

As the self-driving tech industry gets better at checking off the edge cases, I truly hope people stay open to evolving how we see transportation and the benefits it'll bring.

⬐ janoc

Sorry, but nobody has ever shown this safety aspect to be anywhere close to true. But everyone is taking it as a gospel. Musk said it, so it has to be true, right? Never mind that there are no independent data to support it available, except what came from Tesla (who has an obvious vested interest to show that their systems are safe).

It also ignores that there are pretty much no self-driving cars on the market with the exception of some buses running slowly on pre-set lines. Cars like Tesla are not self-driving, their systems are level 2, maybe 3 at best (i.e. human always in the loop, ready to take over immediately). Then there is Waymo who is attempting for level 4, but those are still research prototypes and not commercially sold vehicles. Same with Uber. And both still have tons of issues.

So based on what are we concluding that this is going to be safer, exactly?

Given how poorly regulated the automotive electronics industry is, how opaque all those boxes are and how fierce resistance are the manufacturers putting up to even attempts to have these things opened up for inspection and certification (similar to airline industry), this is a completely ridiculous stance to take.

Especially in the light of recent scandals like Toyota acceleration problems, Tesla's autopilot crashes, Dieselgate, cars that can be hacked and disabled through their infotainment systems (Jeep ...) etc. - all of these were essentially software and electronics issues, often in much simpler systems than safety critical ones. Companies are unable to get even something like a car radio right and you want to trust them that their car won't kill you because of a memory overflow somewhere or camera misreading a stop sign? Only because a manufacturer makes such claim, with little to no verifiable evidence?

Is this the manufacturing culture that is supposed to solve the safety problems on the road? Good luck with that. That is wishful thinking while ignoring both the evidence and past industrial experience.

Even that airline industry which is much more heavily regulated, inspected and supervised than anything automotive has large problems - just look at the 737 MAX debacle.

⬐ mothsonasloth

Well said, cars are becoming closed source both at hardware and software levels.

⬐ mattlutze

Self-driving features are indeed being demonstrated to be safer. The video was talking about the different levels of self-driving so maybe we can agree to discuss it on that spectrum.

Tesla publishes some information [1]. I think the others do too. (Take your requisite grains of salt, I work at Tesla.) The the self-driving features are helping reduce accident rates.

I get being skeptical of cars driving for you. I think we should also be concerned with the bad choices that fully fledged human drivers make, how emotions they don't control (or situations they put themselves in) can exacerbate those choices, and how often those things result in damage, injury or loss of life.

1: https://www.tesla.com/VehicleSafetyReport

⬐ kick

I agree with Hotz, but self-driving cars are an unambiguous win. Way more people die from car crashes than most ways to die; there's a moral imperative to take away control from humans in that regard.

Dating apps are easy: don't use them. Bars still exist. Voice assistants are pretty easy to not use.

Cars are not, because cars aren't consensual. Kids don't get a choice as to whether or not they get to ride in cars. You take a bike, regulators have made it intentionally dangerous for people on bicycles so cars could continue to rule the world. Go for a walk and most sidewalks are in terrible condition outside of a few areas in the United States. Public transit has been systematically defunded, partially because of cars. In America, most people have no choice.

Automate away, there's plenty to think about.

⬐ mothsonasloth

>Way more people die from car crashes than most ways to die; there's a moral imperative to take away control from humans in that regard.

How many self driving cars are on the road vs. non autonomous cars on the road?

As a software developer, we are taught to look at the root cause. In your case you are saying the human is the root cause but that is too shallow.

The root cause, is why a human was in that situation in the first place - stress, crazy work hours, car maintenance etc.

⬐ ribs

So, just solve stress and crazy working hours. Ha

⬐ kick

The root cause, is why a human was in that situation in the first place - stress, crazy work hours, car maintenance etc.

All three of these are solved with self-driving cars. The human is the least-manageable part of the equation. Take out the human, stress doesn't matter. Them drinking doesn't matter. Them having crazy work hours or whatever excuse they'll use doesn't matter. With proper self-driving cars, them being conscious eventually doesn't matter.

This is one of the few problems that can actually be solved with a mostly-technical (including law as tech, because law is programming) solution: take away control of vehicles from humans.

The root cause is that humans are not good at handling motor vehicles going at high speeds repetitively on average. There are too many factors to cover. It's 40,000 people dead and millions injured in the United States every year over something that can be mitigated.

⬐ mothsonasloth

Ok my guy, lets see how this rolls (excuse the pun) in 20 years from now when my Bezos Mk2 hydrocell vehicle running Microsoft Car SP3 has a glitch in its LIDAR driver...

⬐ kick

Congratulations, you picked proprietary software and bet on LIDAR, then died because of those choices. The death count is still significantly lower than it was in 2020, and life is considered better for most.

For a serious answer: even if the life-threatening crash-caused-by-glitch rate is 1/10,000, which is unlikely on its own, that's still giving you better results than human drivers, where the rate of death (keep in mind, deaths, not injuries) is around 12/100,000 drivers.

⬐ axegon_

I can't watch the video atm but I'm incredibly perplexed as to why the hype around self driving cars began in the first place. To me, even if it does become a common technology at some point, it's nothing more than a "hey-check-this-cool-thing" gimmick. For end consumers anyway. If my car has full self driving capabilities, I can't see myself using it more than once every 3 years.

⬐ kevingadd

The idea of not having to drive to get somewhere is really appealing if it doesn't involve having to wait at a bus stop and sit next to a stranger on the bus. People in many environments - especially silicon valley - spend a ton of time commuting every day, either in their own car, on a corporate shuttle or on public transit. So if you tell them they can sit in the privacy of their own car while it takes them to/from work and relax, you have their attention.

Of course, that doesn't mean it's actually a good idea and it doesn't mean it's possible - but it's a seductive idea if you're someone who hates commuting, and many people really hate it because it eats up a chunk of their life.

⬐ baby

I'm one of the people commuting 3 hours per day in the bay.

I don't know what's the solution but right now the problem with commuting with a shuttle is not solved by replacing it with a car, the roads are super bumpy and the traffic is bad.

What I wish was for a subway to take me from SF to south bay. No traffic, stable, fast.

⬐ nullc

No snark intended, but 3 hours per day is a tremendous expenditure of your time... is that really preferable to working close to where you live or living close to where you work?

Other than a brief period where I was commuting between Florida and DC I've never lived more than a 15 minute drive from my office (or worked from home)... I can't imagine sinking 3 hours a day into a commute. There are plenty of decisions in my life that I'd make differently in a redo, but the commute distance decisions are not among them.

⬐ kevingadd

It's prohibitively expensive - if possible at all - to live close to the main campus of a company like Google or Facebook. If you're lucky your commute is maybe 30 minutes, if you're unlucky it's much longer. Self-inflicted harm, perhaps, but if you're already working there cutting down your commute basically means you need to quit your job because most teams won't let you go full/heavy WFH.

⬐ nullc

Up until a couple years ago I used to live within a very short commute from the google campus. My rent was ~$2000/month for a two bedroom apartment in Mountain View. I can't imagine that SF is much cheaper.

⬐ baby

Can you really get this in MV now? There's no way you would even get a one bed room with this money in SF.

⬐ nullc

Yes. Older buildings, nothing fancy. Availability of 2br seems kinda tight so you won't get a great price if you're in a hurry.

Sunnyvale is even less expensive.

⬐ tim333

Self driving shuttles / taxis could work well with fast rail for that sort of journey. Not sure the exact situation in the bay area.

⬐ kevingadd

It's really depressing that for a short time period, it was a real possibility that BART+Caltrain would cover the vast majority of the SF bay area compared to what it does now. So much wasted potential because people weren't willing to eat some marginal costs then to lay down extra lines. We're slowly expanding it now, of course, but decades too late.

⬐ colordrops

> If my car has full self driving capabilities, I can't see myself using it more than once every 3 years.

Huh? Why?

⬐ discordance

Perhaps there’s a bias for it because too many software engineers waste too much time in traffic on the 101 going to/from the city and their SV office.

⬐ jmiskovic

As much as I'd like self-driving tech, the price seems too high. High resolution urban area maps of that need constant updating? Built by and controlled by multi-national IT corporations known for no regard for privacy? How about we pour that research and investments into bettering the public transportation? Would be good enough for most people.

⬐ alexisread

While we differ in execution, this blog is really good at explaining a new form of transport which I'd venture is better than self driving cars:

https://openprtspecs.blogspot.com/2015/02/designing-prt.html...

⬐ kevingadd

A massive barrier to this is that people generally seem unhappy with the fact that public transit means they need to sit next to other members of the public. Rightly so, to a degree - our fellow humans can be quite rude or unsightly or dangerous at times, and most of us have probably done some shitty things in public by accident a few times. So if you offer them a pipe-dream of private automated transit instead, that's a lot more appealing than dropping 8 figures on slight improvements to BART or their local buses.

The area I grew up in had a light rail network that was supposed to expand to my community college within 3 years, and in practice it took over 10. One source of delay was that communities rejected the expansion, and one reason among others was a claim that Local Crime would 'ride the rails' to their previously quiet neighborhoods. Sometimes people don't want to share space with their neighbors.

⬐ baby

the failure of public transport in the US is not a good argument against public transport in general due to its large success elsewhere in the world.

⬐ kevingadd

I am 110% in support of public transport expansion and not especially fond of self-driving cars as an alternative, just trying to express why I find people in my area (and industry) tend to prefer self-driving to transit.

⬐ hef19898

It is also way more sexy. And it attracts way more VC money as well. A well run public transport system is a nice thing to have. It can still be optimized, so.

⬐ tyfon

We have excellent public transportation where I live (Norway), but after I got kids I needed a car. It's not just about public transportation being a success.

We try to use public transportation as much as possible but sometimes it just isn't possible. I can't carry 4 bags of groceries on a public bus while having two kids in tow after an 8 hour work day.

The focus should be making the cars pollute less not remove them (except in city centres where they do not belong).

⬐ mattlutze

Having a bunch of people and things to move around is a good reason for a vehicle, because you're using it. Being a single occupant in a midsized something filling up roads is what folks are generally concerned with first.

Congestion is a big issue and until cars are 0 emissions, putting more on the road will bring more pollution. It will also mean building more roads, which further increases our environmental impact.

Best to both improve that through regulation and industry response, and actively work on expanding multi-modal mass transit options. The goal should still be fewer cars overall along with fantastically desirable other transport options -- we should (in a good way) want you to be disappointed when you have to get in your car instead of taking the mass transit options.

⬐ chrisco255

No, it has to do with the fact that public transit is synchronous and personal transit is asynchronous. For me to travel via public transit, I have to follow their schedule, their routes, possibly hop between 5 different buses to get to my destination (work) and then somehow take 3 other buses and a train to pick up my kid, another couple of hops to pick up groceries, then drop my kid off at ball practice, then pick them back up again after running a few other errands. Doing this in a suburban or even moderately dense urban environment via public transit is a giant pain in the ass, if it's even possible at all. There's slippage in time for every stop a bus needs to make and every route change that needs to occur and for every minute spent waiting on a bus or train to arrive. This is why we built up around the automobile in America. It's not that it's pointless and it's not that we hate other people. It's efficient.

⬐ alexisread

While we differ in execution, this blog is really good at explaining a new form of transport which I'd venture is better than self driving cars:

https://openprtspecs.blogspot.com/2015/02/designing-prt.html...

⬐ hcs

Funny, I tend to think about it the other way: Mass transit is asynchronous, because you wait and something else comes 'round to service you (eventually), and then you wait to get there. Whereas with a personal vehicle, you're engaged in operating it the whole way through, and can't do anything else, though you get to start going at any time.

⬐ melbourne_mat

What is it with Americans and their loathing of one another? What is the problem with sitting next to a stranger? Maybe it's time for a spoonful of concrete.

⬐ pvorb

If more wealthy people used public transportation, the percentage of shady people on public transportation would go down.

⬐ kevingadd

"shadiness" has nothing to do with wealth. Beyond a point you can't really tell whether someone is wealthy by looking at them on the bus. You can tell whether or not they spend extravagantly on fancy clothes or accessories, but you have no way to know whether the person you're looking at is a millionaire who just believes in dressing casually. You could also be looking at someone who's been unemployed for two weeks and is looking their best because they have a job interview, so they look wealthier than they actually are. And none of that has much to do with whether a person will seem "dangerous" to the average transit rider.

⬐ pvorb

You are right. Sorry for my bad and disrespectful wording.

My point was more about that if the percentage of people who currently use their car to get to work will go by public transportation instead, it will feel less unsafe, since there are many other people around.

⬐ mattlutze

The barrier to expansion is that suburban sprawly cities have been designed around cars, shared transportation under-developed, and wealthy folks are scared of being too close to less wealthy folks, who they think will try and steal their stuff or cough on them.

Relatively rich, conservative suburban folks are going to make choices to disadvantage less-wealthy people (like those that are historically going to need public transport and can't afford a car, or a second at least). It's easy to raise a group of people's outrage through scare rhetoric; that's why we the US has it's current administration. It's also not a good reason to just say "oh well buy more cars".

⬐ tim333

Thing is we've had investing money into public transport for a century and things don't change much - London's busses and tubes work much the same as decades ago. Self driving has the promise in the future to massively change things and partly sort the million annual road deaths.

Like I keep a car out in Hertfordshire to get around there because it's the only practical way really in semi rural areas - the busses are rubbish and taxis too expensive. If I could get a £3 driverless electric thing rather than a £20 human driven petrol taxi I'd probably do that and dump the car.

⬐ alexisread

Yep, and if you stick that taxi on a rail hung from lamp posts, then you don't need to worry about batteries, other transport interactions. You can run it straight into warehouses to deliver a pallet load of goods point to point, up the side of buildings eg. Home to 4th floor, run it at 200mph in the open, overnight sleepers for long distances, and if you're rich then pay for some track to your door.

⬐ contrgm

Wouldnt a full self driving solution solve its own problems? I.e self driving cars can perform the task of constantly updating its urban map?

⬐ chkaloon

I don't see how Levels 4 and 5 can happen without a massive civil engineering effort on the road design and construction side of the equation. Everyone talks about the work being done on the cars. What about an effort to better standardize the roads, markings, maintenance, etc.? Until that happens, we will never get there IMO.

⬐ IXxXI

Are americans really dumb enough to kill off silicon valley. One of their primary sources of jobs, wealth and standard of living. Tell me its not true.

⬐ bsder

This is a very thinly disguised ad/puff-piece for Comma.

⬐ throwjaway5

Would you be more okay with it if it was Musk talking about Tesla, or Krafcik talking about Waymo?

⬐ oh_sigh

There's nothing thinly disguised about musk talking about tesla

⬐ RijilV

s/talking about tesla//

⬐ bsder

No.

The problem is that this is a puff piece that cloaks under the guise of being "controversial" and "hard-hitting".

In reality, it is an ad with negligible technical content.

⬐ pauljurczak

He is absolutely right that driver monitoring is a necessary component of any SAE Level 2 and above. It will remain so until capable Level 4 and 5 systems are deployed and verified by billions of low accident rate miles in real world scenarios.

I will risk predicting the future here: Tesla will have to add a robust driver monitoring system (beyond steering wheel torque sensor) to their cars with "autopilot" enabled.

⬐ rr98x

If that happened, it'd be analogous to iMessages going open-source. Even if most people pay attention in Autopilot mode, the current system is already somewhat demanding. If it were greater, I think it'd lose its appeal.