This Hacker Is Teaching a Smartphone App to Drive Your Car

The visionary hacker on how he plans to "solve A.I." and why he thinks this will be a "decade of decentralization."
------------------
Subscribe to our KZbin channel: kzbin.info?sub_...
Like us on Facebook: / reason.magazine
Follow us on Twitter: / reason
Reason is the planet's leading source of news, politics, and culture from a libertarian perspective. Go to reason.com for a point of view you won't get from legacy media and old left-right opinion magazines.
----------------
George Hotz, the hacker and entrepreneur who as a teenager became famous as the first person to jailbreak the iPhone and then the PlayStation 3, is determined to build a human-level artificial general intelligence. As a first step, he's racing to build the world's first fully self-driving car system.
Hotz's company, comma.ai, which has raised $8.1 million, is competing against giants like Google offshoot Waymo, Amazon-owned Zoox, and General Motors project Cruise. Those efforts rely on custom vehicles with expensive sensor arrays such as LIDAR-light detection and ranging-to identify surrounding objects and people, navigating a pre-mapped path. Hotz says his better-financed competitors have misled the public with hype about what their products can do and on what time scale.
But comma.ai takes a radically different approach, using vision alone to analyze the road ahead and drive as a human would. Its latest product, the comma three, runs on a smartphone processor and plugs into most new cars, taking over the built-in steering, gas, and brake systems. The company's open-source software, openpilot, uses artificial intelligence (A.I.) to ingest the comma three's video feeds and predict where to drive in real time.
While Tesla is also moving toward a vision-only system, Hotz says comma.ai's approach is unique: The latest version of openpilot features the self-driving industry's first "end-to-end" architecture for lane keeping, which means that it won't explicitly identify objects and people or rely on road features such as lane lines. Instead, the software analyzes a scene holistically and determines the most likely path a human would drive.
Hotz says only this approach to computer vision and self-driving can lead to the development of robots that can cook, clean, and do manual labor-and ultimately to an A.I. system that can do everything a human can.
According to Hotz, comma.ai's commitment to open-source software also represents a philosophical split with companies like Waymo that are building closed systems with centralized control. In the self-driving space and the larger tech industry, Hotz is optimistic this will be a "decade of decentralization," and predicts that the 2020s will see distributed systems powered by blockchain technology-including cryptocurrencies like bitcoin-overtake corporate giants like Visa in both trust and value to consumers.
Produced, written, and edited by Justin Monticello. Director of photography Dillon Mortensen. Camera operator Elvis Leon. Audio production by Ian Keyser.
Music: The City of Hope by Borrtex.

[@jaredhead9503]

An interviewer with knowledge about the topic being discussed! Astounding!

[@ericweis9771]

Yes journalism is not completely dead!

[@alimohammed7100]

@@ericweis9771 I agree! he has been following this story for over 4 years

[@oryoruk]

I applaud the Reason interviewer Justin Monticello, all excellent questions!

[@watchingvideos9871]

Exactly what I was thinking. What a brilliant interviewer.

[@beep_]

yep i think this is their third interview together over the years, each one is great and well worth a watch

[@circuitsandcigars1278]

And no use of UMMMMMMMMMM which tells me a speaker was ill prepared and didn’t practice talking ahead of time

[@rubenbiskupec8442]

I have to a agree, I am big fan of Lex Fridman but I have to admit this conversation was even better! I can't wait for the next interview in a couple of years

[@IvelLeCog]

God damn refreshing is what it is

[@tipoomaster]

Beard Guy always impresses me with his interviews, it's never like others where it feels like they're missing a point a smart person made.

[@martinlutherkingjr.5582]

Every answer he gives is deeper than expected.

[@sanjacobs6261]

And yet so well distilled

[@DoisKoh]

Lidar is off the mark though. Though I'm sure he understands and is just being hyperbolic. If, and yes it's a very big if... lidar did become cheap enough, it definitely will benefit self-driving. In certain situations such as extreme snow/fog, lidar will provide critical (life-saving) info that is simply impossible to achieve with pure cameras (or at least, you can argue that you would be able to drive much faster, safely in such conditions).

[@zofe]

@@DoisKoh When a human cannot see - then better don't drive into the abyss.

[@9jmorrison]

I think de-centralization is important due to the nature of humanity, from corporation to politics it seems to be run by sociopaths.

[@multiverse7797]

Precisely!

[@nobodynoone2500]

only when "de-centralization" doesn't actually mean consolidation on the hands of a select few hosts and corporations.

[@jjs8426]

Tell that to the new world order

[@JohnDoe-fz5cz]

business isn't run by sociopaths, but politics is.

[@bloodraven2704]

@@JohnDoe-fz5cz bullshit, many of psychos or sociopath are in business too. Just start to interact with rich owners of business. They are in every field of the life : science, business, politics. These people love to progress in something : power, money or just achievements. George Hotz, by the way, has narcissistic characteristics.

[@PatrickHealy]

This may be the best Reason video I've seen this year.

[@supernoob17]

it should be illegal to make a documentary about geohot without mentioning the time he got sued by sony and responded with a diss track

[@richctv]

There are multiple reasons you don't hear much about George Hotz. He's rather visceral and to the point on his competition. So very little lip service in that he doesn't tow anyones line. No Woke talking points. Q:What about Universal Basic Income? "Where is this Free Money coming from? Who has the authority to give Free Money? Basically a pariah in Silicon Valley.

[@trod5902]

and what I love about it is that he does so in such a nonpolitical fashion, its purely pragmatic.

[@Ben-rz9cf]

He's also pretty ignorant in his own field. Anyone working in computer vision knows the importance of feature verification in deep learning sample sets and he's outright saying he doesn't care about the why and even trashing Teslas label driven approach as if its somehow inferior for having that extra verification layer. You absolutely need that if safety or reliability is at all a concern for you. You want your deep learning to be as transparent as possible to your engineers, lest the computer come to some kind of obscure decision you don't understand and can't unpack until you build an additional feature layer to actually analyze it. And when it comes to self driving cars, thats literally putting people's lives at risk, there is a reason Tesla insists on that extra layer and its not just for the regulators, its for their own liability as well. They want to be able to say at any point if something went wrong they know exactly what happened and that they accounted for it as best they could. And trashing lidar? Seriously? It's one thing to say it's expensive but to say it like its unnecessary or inferior is just amateurish. At the very least you need some kind of TOF or stereo vision, single camera CV is deeply inferior to depth based methods and he should know that.

[@HsenagNarawseramap]

@@Ben-rz9cf 1. He assumes no liability so he doesn’t care about ah intermediate layer. 2. Karpathy has moved away from Lidar as well.

[@JohnDoe-fz5cz]

toe anyones line. as in "toe" the mark.

[@nightking1863]

@@Ben-rz9cf lidar is for positioning not for vision you're not very knowledgable on this topic, also comma is lvl2 forever so no liability

[@agustinpizarro]

This guy is freaking incredible. I would have needed days to think about some questions. And some more time to try to explain my answers so easily and clearly.

[@RyeOnHam]

If Musk, Zuck, and Jobs had a three-way love child.

[@vikramsarabhai1]

For some reason all these geniuses come from the same group.

[@xythiera7255]

@@vikramsarabhai1 But the arnt geniuses at all . The made a product that made the lives of peopl easyer so it soled . But the arnt super intellegend peopl .

[@1voluntaryist]

@@xythiera7255 Maybe not by your definition of "super intelligent". What do you call a person who can look at the same data as everyone else, draw a unique conclusion that no one else agrees with, and be correct? How does that happen? What gives one person the ability to see reality better than a billion others? And would you call that person?

[@trod5902]

George always blows my mind when I listen to him. Great interview

[@iclick4122]

And his best, authentic thinking tends to be correct.

[@petabb]

First time I see George at his best all thanks to Beard Guy. He is an awesome interviewer.

[@steve-g-]

Absolutely excellent interviewer.

[@Vorpal_Wit]

Yeah. I hope Gilespe is taking notes. I dont recall a single overtalk or interruption in this interview.

[@mcresearch]

An excellent interview - questions were seamless and unintrusive. Well done to both.

[@shir0923]

Glad to see Hotz again on Reason, been a few years since that first special.

[@PoopiteeScoop]

Holy shit-a knowledgeable interviewer on a topic I’m passionate about! Man, I really enjoyed this. Wonderful interview! 👏👏

[@MiscMitz]

"I don't care how you got to the right answer, just that you got the right answer" Perfect

[@PhyloGenesis]

That's a terrible quote... It's ridiculous to say that because if it were that simple we'd have solved all these problems already; my Tetris AI would have been superhuman instead of not working at all for example. The reason we have all these intermediate layers of, say, object recognition or sentence parsing, is that just evolving the AI directly to the holistic solution is extremely hard and almost all attempts have failed. That's why tons of types of neural networks exist. We use smaller steps like evolve it to recognize a lane line, now to understand a pedestrian, etc., so that we can break down the problem and build toward the whole solution. Also, understanding why it isn't working is critical to human-problem solving like me trying to fix the training process so that it doesn't hit that cone next time. Just jumping straight to the end has been tried countless times in countless areas and it doesn't work. If that made sense, why is he working on driving AI and not just going straight to AGI? Same reason Tesla is working on perception instead of driving. EXCEPT, this guy has actually been meaningfully successful with this problem so he gets a pass to say it. But as a general way that most people should operate, it's horrible.

[@MiscMitz]

@@PhyloGenesis I was more speaking on how schools are teaching things like math currently. They don't care about the right answer. Lol

[@MiscMitz]

I tried helping my 9 yo with math and I couldn't. 😆 🤣

[@PhyloGenesis]

​@@MiscMitz Oh lol. Well in that case I still totally disagree. :p The way I was taught is completely useless now thanks to calculators and phones. Fortunately, I naturally think in a general problem solving way that I applied to math that made it super easy, and that has remained widely useful in almost every aspect of life. That way of thinking is also how everyone else I know that also found it easy said they think. My limited understanding of Common Core is that that's basically what it teaches. So that's great that you know this one answer by luck or some other means, but you didn't actually demonstrate that you learned what they are trying to teach.

[@MiscMitz]

@@PhyloGenesis good point. Thank you

[@FukU2222]

Truly an underrated gem and wealth of knowledge... Bless George

[@forntoh]

The best interview we've watched in years 😭😭😭

[@tipoomaster]

George "If you have two Shitcoins" Hotz

[@willstikken5619]

I think my biggest fear now is someone creating an AI that learns how to govern by watching American politicians...

[@rickjames5998]

Or Chinese Government!!!

[@willstikken5619]

@@rickjames5998 If you followed the approach comma is taking then it wont matter much. The differences are mostly a matter of degrees and an artificial venner of choice. Many of their actions and the direcitons these countries are oging are aligned.

[@proehm]

So, it has to learn right from wrong and how to not be a megalomaniac first...

[@ezanchi5422]

Awesome interview, really great questions being asked, really good topics, you let the interviewed reply and did not interrupt him even a single time -which nowadays even TV interviewers sometimes do-, good video editing... New subscriber here

[@TheMichaelMove]

So cool! But I’ve been surprised by how many free market lovers seem to be anti this kind of tech.

[@macdietz]

Why are you equating free market economics with potentially dangerous technology..?

[@TheMichaelMove]

@@macdietz uh because that’s what it is? You have the freedom to use it or not.

[@JamesTheAxeThrower]

@@TheMichaelMove we might not have that freedom in the future.... that’s why it is dangerous.

[@aussie2uGA]

Mainly because history shows weaponization of it.

[@TheMichaelMove]

@@JamesTheAxeThrower so the government should arrest the inventors?

[@celebrationsbridal]

I think I'm smarter for seeing this interview, but it sure don't feel like it. Damn, check out the brain on that guy.

[@Lets-Go-Family]

Hotz is an underrated genius of our time.

[@JD2jr.]

less "underrated", more "unpromoted".

[@garystinten9339]

Imagine if he had Elon levels of money

[@Lets-Go-Family]

@@garystinten9339 He probably will.

[@nunyabuisness21mill]

Yeah clap, lol cracks me up every time. I think of Jeb bush “please clap” lol

[@Emotionengne]

18:53 "The only difference between human and the birds is that humans think they are special".. He chalks it off as a flipping answer but that's the most profound and true statement. He is absolutely right about that. Hence we create problems and create solutions to solve those problems which creates even more problems and we go on and on. Actually, we are not that different from the machines, just like the birds.

[@carrocesta]

brilliant interview, thanks ReasonTV!!

[@promenade-uncouth]

Really good interviewing! Definitely asking good questions and not just agreeing with everything being said.

[@Smalltime]

Very good interview! Hate the title thought, it’s rather belittling in regards to what Mr. Hotz is doing.

[@LoganLeGrand]

They changed it and it doesn’t even make sense anymore. The Comma Three isn’t a Smartphone and OpenPilot isn’t an app.

[@ShanyGolan]

Hotz is down to earth smarter than the average guy, yet speak in simplicity

[@proehm]

A good sign sign of real intelligence, as opposed to a lot of education, is being able to explain things so that everyone in the room can understand.

[@samuelkipyegon]

Great interview. Great questions and insightful answers. I only wish it were longer.

[@glenkru]

Excellent interview. Thanks guys. Always interesting to hear George's thoughts on a wide range of topics.

[@ce1tzu]

Great interview, interviewer and interviewee!

[@jasonrojas26]

A fantastic interview on both sides. Well done

[@emera1750]

So glad he’s still around and innovating on his craft Inspired me so much back in the home brew days so great to see how far he’s going

[@Ramirez4565]

Great video. remember seeing him back in the day. excellent job

[@offgridas]

Pretty impressed with the quality of conversation here. Good questions, interesting answers

[@heresthething5379]

"Last time we talked, it was 2 months before the pandemic hit". Eyebrow raise 🤨 "Or two months before the lockdowns" Nods head 😂

[@kamatihasheela2995]

This guy's intelligence is on another level. He his able to provide a convergence solution to each of the questions posted to him.

[@multiverse7797]

Great interview! Very interesting exchange.

[@BonanzaPilot]

we got our comma 3 quicker than I thought we would. Waiting for the mount to cure. Excited to see how it does compared to EAP in our Tesla

[@bhoovis]

This is an excellent and inspiring interview...nice work.

[@sebleaf8433]

Excellent questions from the interviewer.

[@eliasb8]

Excellent questions and excellent answers. What a great interview!

[@sfaxo]

Love my comma 2 and recently got my comma 3. Love this tech. I've been using it for almost a year and it's such a better driving experience.

[@jordansumitomo2357]

Let get this video to a million views

[@ericweis9771]

Fantastic guest and great interview thank you

[@CastroBlues]

When we get to a point where AI can drive by itself in NYC during rush hour I'll be sold.

[@tann_man]

LMAO living in NYC in 2021+

[@PhyloGenesis]

I think his already can... not sure though.

[@Teslavangelist]

Really great depth questions by Justin

[@israelhydroponics8659]

Great interview, really.

[@bobrein6368]

Such a well done interview 🙂 I really liked his explanation that humans are a similar black box.

[@ShanyGolan]

I love the dream machine in the background.

[@RichardBronosky]

You got to respect a man who has a Unifi Dream Machine prominently displayed in his lab.

[@haydenmacfarlane7194]

Awesome interview!

[@CJ-eo2xz]

1:49 his eye roll is perfect

[@rockdamic]

Remember when there actually was a time when we would see interviews like this on TV?

[@DreamHound]

Its funny only yesterday I saw the previous two interviews and today they release this. 2 things have evolved from watching all three talks, one is my perspective on self driving and 2 on the interviewer's experience in asking question you can just see experience kick in better.

[@chancerobinson5112]

Mr. Hotz is unstuck from time. He’s sure it will happen and won’t be pinned down as to when, because he doesn’t care! He believes it into reality. Very powerful.

[@NeverTalkToCops1]

Believing doesn't make stuff happen.

[@thomasthemarstrain2141]

INTERVIEWER DESERVES A RAISE!!

[@Syncopia]

Excellent interview and interviewer.

[@TaylorMade3906]

Fantastic interview

[@mafa5534]

Really decent interview!

[@TamasVarga-VatartPhoto]

Amazing content. I am a big fun of Georges, he is a genius and truly visionary. He is also into economy, philosophy etc, I love it. Interviewer also did a very good job. Good questions.

[@m3po22]

19:00 Different layers of adaptation

[@abramgaller2037]

George Hotz is quite brilliant .

[@makelife9661]

Great interviewer, great guest!

[@INTOTHEFOLD]

This was fantastic. Well done.

[@kipling1957]

George has been notably quiet following Tesla’s AI day.

[@garystinten9339]

Cause he's hard at work making it with an n64

[@thems_the_brakes]

@@garystinten9339 😂

[@_dinesh]

Wow fantastic interviewer.... all great questions!

[@sdmarlow3926]

11:09 is the key take-away. ML deals with last mile issues (sorry for driving metaphor; actually applies to any domain specific challenge) without ever building the hundreds of miles of infrastructure required to really solve last mile issues. AI, real AI, AGI, whatever you want to call it, is about THE system that can deal with playing games, driving, engaging in hours-long conversations, etc.

[@joelchong12]

Awesome video!

[@switzerland3696]

At 1:00 when you introduce the comma 3, you are actually show production footage of the old comma 2 lol.

[@jimwall2291]

The shitcoin analogy got me. The whole interview was great but that was really good.

[@samuelt.5215]

A very brilliant mind!

[@EMO_alpha]

This was the best video on reason i have seen in forever lol

[@pocki892]

Yezz I feel twice as smart now!

[@jasonjasonjasonjasonjason]

George is making force field wristbands

[@damononthehill]

I've been following this technology for a few years, it's incredible. Years ahead of the big players

[@Gvue003]

Why is no one mention this to other platforms?

[@Kefka.]

But current Tesla cars basically are self driving cars. He use to brag about taking less long of a time compared to competitors. And some of his comparisons are strange, we have 20 years of driving data, 20 year olds can drive? "So there you go." What!? More interviews I hear with him make me more and more skeptical.

[@doom2avatar]

Man is a quack. If you actually look up the results from his classifier it cant even tell read apart from sidewalk.

[@Kefka.]

@@doom2avatar There are so many strange things he says and dismisses out of hand. If the car stops every-time at a stop sign, I don't care if its because of the sign, the white line or something else the point is it stops. This is very specious reasoning, of course you want to know. What happens if the only reason is the white line and in some rare places there is no white line,etc. Also he point out the diminishing returns of more and more data, while technically true if the more and more data is massive amounts of data that mitigates some of those diminishing returns. And some returns is still meaningful.

[@Kefka.]

@@user-io4sr7vg1v If he actually produces a good product good for him (and us). But a lot of his answers seem like evasions or something that might sound surfaces level reasonable but less so more you look into it. This is the 3rd long form interview I have seen of his over the years plus some articles and he does not come across well to me.

[@Kefka.]

@@user-io4sr7vg1v This is the 3rd long form interview I have watched of his.

[@JoonPark1]

I didn't know George Hotz was so intelligent... even with regards to things unrelated to programming and AI

[@wulfboy_95]

A lot of new flagship phones are shipping with Tensor Processing Units nowadays, so stuff like this is pretty much inevitable.

[@startek119]

When are reporters going to ditch the lidar questions.

[@peberdah]

Super interview, Georges is a genius and he definitely path the way to Net application design. Decentralization and multiple copy AI must be core as hearth beat, a filter layer to personalize result, from time to time a super computer boost injected in the network.

[@fantinuum1040]

I would have to correct George on one thing. Mark was not the first to drop out of Harvard and start a company. Bill Gates was the first one to do that, along with Paul Allen. And there were others, too, in that era. Steve Wozniak (Apple), Larry Ellison (Oracle), etc.

[@OneMarko]

Isn't there a paradox in trying to learn from people to get better ai and become greater over time, where people and corporations may fail in some way? Will ai not learn to also fail some way if it tries to mimick human actions?

[@newagain9964]

I want to see it on a course. Having to respond to actual challenges. I bet it fails 90 out of 100 attempts.

[@FranciscoRamirez-gb6zc]

He is right about the taxes… that’s why people are leaving California and NYC

[@mentalgymnastics6384]

Terminator starts out as a chauffeur interesting.

[@davidstrong7854]

People who are geniuses just operate on another level. Proof that’s life not fair.

[@NikeshAmin]

Is that a Unifi Dream Machine in the background?

[@Ebacherville]

I use Open pilot every time i drive, drive better than most humans on the road.

[@proehm]

I don't have a car that's capable, but heard someone say "It doesn't take the responsibility of driving away from you, but allows you to eat your sandwich or fool with the entertainment system without becoming a menace." That would have to help.

[@dimitribiemond]

This interviewer is extremely good

[@vonbayernDE]

Gold.

[@TurntBucket]

6:07 he looks very concerned. lmao, thats a little troubling.

[@tyclips7143]

Love how they don’t want to say the name Tesla.. But mentions GM 😂😂😂

[@jamesyjamesik715]

29:12 the way he claps his hands and nearly misses because he was deep in thought extrapolating what "competition for governments" means to him made me a little nervous I have to say. Especially with that hidden smile as he turned his head 😭😭😭🤣🤣. What is he up to in that black box of his 😂😂. Good interview

[@theketixxx3835]

07:33 my math teacher has a different philosophy

[@QuadCloudNine]

man Reason loves interviewing this guy.

[@JD2jr.]

And we love that they do.

[@jamesclarity1077]

Always amazing