Tesla vs Waymo: The Race to Level 4 | Boris Sofman and Lex Fridman

Lex Fridman Podcast full episode: • Boris Sofman: Waymo, C...
Please support this podcast by checking out our sponsors:
- LMNT: drinkLMNT.com/lex to get free sample pack
- Athletic Greens: athleticgreens.com/lex and use code LEX to get 1 month of fish oil
- ROKA: roka.com/ and use code LEX to get 20% off your first order
- Indeed: indeed.com/lex to get $75 credit
- BetterHelp: betterhelp.com/lex to get 10% off
GUEST BIO:
Boris Sofman is the Senior Director Of Engineering and Head of Trucking at Waymo, formerly the Google Self-Driving Car project. He was also the CEO and co-founder of Anki, a home robotics company.
PODCAST INFO:
Podcast website: lexfridman.com/podcast
Apple Podcasts: apple.co/2lwqZIr
Spotify: spoti.fi/2nEwCF8
RSS: lexfridman.com/feed/podcast/
Full episodes playlist: • Lex Fridman Podcast
Clips playlist: • Lex Fridman Podcast Clips
SOCIAL:
- Twitter: / lexfridman
- LinkedIn: / lexfridman
- Facebook: / lexfridman
- Instagram: / lexfridman
- Medium: / lexfridman
- Reddit: / lexfridman
- Support on Patreon: / lexfridman

[@neelsviljoen1383]

Extremely informative when honest people discuss and give credit in a nuanced way. Never understood the advantages, risk and disadvantages of both Tesla and Waymo so well. Thank you 🙏

[@theni3762]

i honestly find Boris to be way too defensive

[@DanStreiffert]

What will happen when a FSD Tesla encounters a Tesla Bot in a crosswalk? Will they actually communicate?

[@zbyszanna]

Yes, they have a secret light flashing signals.

[@khuo0219]

One honks, the other gives the finger. That is how I imagine it.

[@francescogiacomelli403]

Love this question!

[@remajudge5699]

For them the point when they meet will be actually already the past.

[@Spruce_Goose]

The last sentence is chilling. When it's solved it can be deployed in so many other scenarios that will change our lives forever. I for one welcome our new AI overlords.

[@cabanford]

The sooner the better. Trump, Putin, Boris, et al need to be delegated to the dust bin asap.

[@madsleonardholvik3040]

@@cabanford Interesting. What do you imagine?

[@glencecil293]

Elon has already thought of this. Have you not heard of the tesla bot?

[@derekketcher9154]

are you hoping they have mercy on you? Well played sir

[@lemongavine]

One of those other scenario is the Tesla Bot. Elon said it just makes sense to use the vision FSD computer for a bot.

[@aztronomy7457]

It doesn't matter who gets to level 4 first. The regulators will take so long that it will allow the other to catch up before level 4 can be properly utilized.

[@WesWilliamsonPage]

Love these conversations

[@adithyakaravadi8170]

Those are some very keen observations, Boris! Thanks for sharing them.

[@Mk4twoG]

Comma ai’s openpilot is great too

[@vasumaru242]

They are great but going to L4 is a different ball game. You need much more complexity in terms of self supervised learning.

[@HeatherSpoonheim]

I'm not an Elon fan, but I do think data is more important - there are just so many unusual situations that no programmer is ever going to think of.

[@mrt094]

You can also get better Data by having a better suite of sensors it isn't zero sum.

[@harsimranbansal5355]

@@mrt094 you need scale, more sensors can be helpful but tesla has enough data to show that they aren’t useful.

[@matthewhindson3133]

You get better data by the range of different scenarios that need to be learned “edge cases”. You can have millions of miles of driving data but none of its useful if it’s all the same. The various scenarios and the amount of the same scenarios is key for machine learning. You are more likely to have these scenarios with more driving data, but you also need to be able to sort through the data to find these scenarios. So the more data the more likely for different scenarios the better, but sorting the data is key to better quality.

[@gazlives]

i love this common qualifier many people put "I'm not an Elon fan, but..." that Elon dude is definitely marmite.

[@taylor92493]

@GB it’s weirder when you consider there are fans for any serial killer, but this has to be prefaced to not get shit on by people who care about Elon too much. In this context, their following sentences didn’t even relate to Elon. So even weirder.

[@xXKM4UXx]

As you are actually creating an intelligent being, the arguments shift from what is good and robust engineering to what is the best design philosophy for making an intelligent robot that can perceive and act in the world.

[@jordanbrown2485]

He reminds me of the dude from scrubs lol

[@M.-.D]

Where are the ads? Are you demonetised?

[@glyphsuritos6588]

waymo will probably "win" the race in the shorter term. i see them reaching AD first with the trucking business since they will work within a bounded geolocation, meaning less rooms for mistakes to happen. easier to perfect AD with trucking. and since their business model is B2B, more likely to turn a quicker profit as business are more likely to utilize it. However, tesla will win in the long term, each data they collect will add more to the true sense of AD. and with it the potential for a robo taxi that drives everywhere, coffee delivery with starbucks, food delivery with restaurants, grocer delivery for walmart, etc. tesla has the possibly to expand more into different areas of the business sector, but it will take time to reach that point. it will probably look parabolic, where many will be skeptical about it, thinking it a far long time away, and but once it reaches that point, it immediatly takes off.

[@getstuk87]

Tesla has been at it for 10 years and their product is full of quirks, at the rate their software has improved over the years I don't see them "solving" FSD or hitting the safety margin of the average person when FSD is on in town for at least another 10+ years. Watch waymo videos vs Tesla, it's night and day

[@denzilbuchner6596]

Perhaps LIDRAR is a risk. You can add training wheels on a bicycle and claim that it is the less risky approach because you have less chance to fall, but try mountain biking or road-racing with them. By adding them on, you risk losing the real game.

[@_____case]

They only care about point to point transportation in major metropolitan areas. That's all they need to achieve profitability. LIDAR will only ever benefit the user's experience given their strategy.

[@georgehermans2187]

Are used to be Terrified by AI. Now after the past couple of years I believe that there’s no way for machines to screw things up worse than we have!

[@flashfive12]

The true terror of AI (as with any intelligence whether it be synthetic or human) is the way humans will wield it. From espionage, to terror attacks, to subscription fees, to human development. I personally fear the day quantum computing, becomes more feasible. That combined with nuclear fusion & AI would be unstoppable. It may even solve the issue with AI all together & create the world’s first bottom up AI.

[@christophercullen5256]

I think Waymo will get level 4 first, and this is great for Tesla because they will trail blaze all the legal/regulatory terrain. Even if Tesla gets there 5 years later, the number of autonomy ready vehicles on the road will be incredible. I think people are also discounting how valuable a high level urban 'cruise control' will be to consumers. I don't want a car that will drive itself yet, but I would love one that does most of the work for me.

[@5893MrWilson]

Waymo is already level 4 in small geofenced areas but they just don't scale. Tesla is probably 2 years away from level 4 imo but they are generalized and much more robust in edge cases. Waymo shuts down in the rain for example.

[@harsimranbansal5355]

@@5893MrWilson I think it’ll be quicker, like its really good in California, and if they can make it as good as a human in just California, thats basically level 4.

[@christophercullen5256]

I think the distinction that needs to be considered is 'level 4 capability' and 'level 4 operational'. To sell a vehicle/service that IS level 4 implies an acceptance of liability if that use case fails - > can not fail and hence waymo chasing the autonomy decimal tail and simplifying/limiting their problem. Tesla vehicles will be level 4 capable for a long time before they certify it as such.

[@5893MrWilson]

@@harsimranbansal5355 I hope it does happen quicker. I think it is technically possible in 12-18 months if things go well with dojo, just adding in some extra time to account for delays. Also hope that legislation doesn't slow it down too much.

[@steve-rb9bm]

Somewhere in between

[@davidbrewer2324]

Bro definitely an Alien. Eye tweaking like crazy

[@djbass7672]

Good to see Checo moving on from f1 from time to time

[@gazlives]

waymo is already level 4 in geo fenced area so they will be first on a smallish scale. but their solution is very expensive requiring geofencing etc. all the benefit to the consumer is not available with waymo's solution. robotaxis for example with waymo will not be much different in price from current prices due to the cost of the infrastructure and maintenance of a waymo system.

[@muhlor]

+ what i learnd about Waymos LIDAR solution: It dies the moment it starts raining. Which hasn't changed for years.

[@UncleGreatness]

Let's ask someone thats working on waymo, which is the better system 😉 🤔 No bias here!

[@DeathZeroTolerance]

As an analogy, Waymo seems like NASA's SLS and Boeing's Starliner...vs SpaceX....Tesla and Comma appear to have the winning approach, with the latter having the most realistic expectations in the pursuit of autonomous driving.

[@eanerickson8915]

Drive on a dark rainy night. There is absolutely no way you are going to just use cameras.

[@xploration1437]

Cameras can see more than what your eyes do. Your ignorance is showing.

[@eanerickson8915]

@@xploration1437 Really? There are more registers to see from?

[@xploration1437]

@@eanerickson8915 oh sure. Many.

[@denisrafailov6056]

If you reverse his bullishness of Waymo towards Tesla he would have nailed it 2 years ago. Even at the end “once achieved if you use that technology for similar problems” you mean like Optimus?

[@shuriken4852]

Waymo's tech of using LiDAR with high res maps in a fenced area, combined with the $200K price per car which they cannot mass produce, just cannot scale, while the Tesla approach can with no problems. Plus AI is a game of data and Tesla has order of magnitude more than competitors. Game over!

[@Heaney5551]

LiDAR drops in price every year. The total car price is down to below $100K now. Tesla can scale, but it has to prove its tech can actually work safely first! No point scaling a product that doesn't work.

[@shuriken4852]

@@Heaney5551 More cars gathering data on the Beta each month + DOJO coming online later next year = Exponential progression. Also I can easily explain why I have no confidence in Waymo, it's simply because Google/Alphabet does not either, as they have over $100B of cash on hand and they still went to external investors for Waymo latest capital raise and this is knowing that the experts say that FSD will be worth trillions in the future, so if they had any confidence in their tech, why wouldn't they invest their own money and keep all the future profits, just does not add up.

[@daltonschwentker]

What would happen if people realized they could safely cross a crosswalk without worrying about getting hit by a car? This situation occurs when we realize that an autonomous vehicle is heading our direction and have a high degree of trust that it will stop before hitting us. On the other hand, when we encounter a human driver, uncertainty sets in and we are unsure of whether they are paying attention or not. In a possible future, people will be able to cross the street unimpeded since they know that all autonomous vehicles will stop, but this will result in more frequent stops and longer travel times for those vehicles. How would you resolve this?

[@nilsfrederking62]

I have seen videos of Teslas driving in very bad weather conditions with poor visibility astonishingly well, never saw a video of a Waymo car in heavy rain. I suppose the Lidar will become useless in such scenarios. If you have any such scenario where a Waymo can master it, please share it. It would make sense to establish a signal language with which autonomous cars cold communicate with "other" drivers and pedestrians, a simple already existing one is the flasher to give someone priority.

[@getstuk87]

Poor weather driving is not something you want to trust an autonomous driving model to handle. Way too many factors at play for now. Waymo is so far superior and how well they handle Inner City traffic and obstacles it's not even funny.

[@slopedarmor]

@@getstuk87 how much does one of those waymo cars with all the sensors cost tho? consumer cars will also become driverless, but probably not with 20 lidars. to me teslas path to making lots of money off it seems a bit clearer than waymos path to riches.

[@spyral00]

It's possible that tesla's FSD will never be level 4 or higher, until we have true AGIs in cars. There's an infinity of edge cases that require a deep understanding of the situation and intentions of the other drivers, that is embedded in the human driving process because humans have always been the drivers. Can that be automated with big data algorithms? Maybe. The data certainly is there or will be, but it's a huge gamble. From what I've seen, Tesla FSD still glitches a lot when "vectorizing" and labeling its environment. I guess when that is solved (and I have no doubt it will be) it will already be a huge step forward.

[@jimj2683]

With enough data 99.999999% of the edge cases will be solved. The rest doesn't matter as long as the car is safer than a human driver. Also, remote drivers could take over via starlink and manually drive through those edge cases.

[@glencecil293]

The ending of this clip is funny: "Here's is whats really beautiful about solving the autonomous driving problem, can you imagine what the technology can be used for?"... Elon has already has the answer...Dude, have you heard of the tesla bot?

[@ryananderson6321]

Can Tesla just starting using Microvision lidar already

[@996vtwin2]

We need. to make more strides in AI before atonomous driving becomes commonplacee.

[@adelinaquijano1083]

Tesla before I have no idea.but I learn more and more

[@Woke-Fact-checked]

Can’t see it gaining traction no pun intended

[@LionheartLivin]

Take a risk, intend the pun!!!;)

[@barrymiller99]

Naïve question: Would there be radiation issues if most cars extensively use some form of radar? Naïve issue #2: Today, some cars run out of gas every day while on the road. What will it be like when countless cars run out of electric power every day on the road? Will chaos ensue?

[@lemongavine]

The cars will stop to charge autonomously when needed. Not a big problem to solve. With 300 to 400 miles of range, they shouldn’t need to charge very often. Even if daily, they can do it at night when driving requests are low.

[@Global_Optimization]

Radar uses radio waves. Completely safe. Dangerous radiation has shorter wavelengths than visible light: UV, X-ray and Gamma

[@barrymiller99]

Thank you, Saxon

[@LionheartLivin]

Thanks for asking these questions!!!;)

[@MrBeagus]

Seems like Waymo system is taught how to drive and the Tesla system learns how to drive. The amount of data that the human brain processes while driving is likely quite small. It is how that data is sorted and used. What is most sure to succeed and what is ultimately best... Seems to me that the Tesla way has more risk, but more potential.

[@lemongavine]

A high percentage of human driving is instinct. Yes, there are rules that you must be taught, but beyond that, the rest is basically easy. Machines don’t have instincts so they need to be trained on all of those aspects.

[@kylegrove2165]

This debate is fundamentally flawed. Anyone working in tech manufacturing knows that you leverage off market innovation as a whole. Tesla will simply deploy sensors in future models when it becomes feasible to do so and in the meantime they will have the upper hand in understanding scalability which is everything in production. From a business perspective it's very difficult to see Tesla losing.

[@KeeeeenW]

Waymo: we are building and burning cash for Level 4 from the beginning. Tesla: we promised Level 5 from the beginning, got paid since Level 1, and got safety drivers for free. Thank you Tesla owners for paying for simulation, data collection, infrastructure, DL model training and development, etc. You will have FSD next year.

[@ericmichel3857]

Or the year after that for sure! They just need a bit more of your time and money. :)

[@LionheartLivin]

My guess is FSD 12-13 will be comparable to the average driver!!!;)

[@KeeeeenW]

@@LionheartLivin Yeah! The recent FSD beta looks very encouraging.

[@Jeddin]

Why the focus on who is going to “win”? If Waymo can reach level 4 1 year, 2 years, 5 years before Tesla will that really matter to Tesla as a business? Will Tesla go out of business because they are not the first to achieve fully autonomous driving? Clearly people will still buy Tesla cars because they are great electric vehicles and Tesla with achieve Level 4 whenever it achieves Level 4 and it doesn’t matter if Waymo gets there first

[@rpcruz]

Yes, people like to talk about countries winning and losing, or companies winning or losing. But it's not like there is a war lol.

[@lizs821]

Tesla’s approach is generalized and eventually will be able to adapt to practically any situation in any environment. Any company, such as Waymo, that uses Lidar and is reliant on high definition maps Is doomed. They essentially are on rails and can’t interact with the ever-changing and dynamic driving scenarios. Generalized real world AI is the more difficult and time consuming solution but inevitably will be the winner by a longshot. Just ask Teslabot in 10 years. It will tell you so. 🤖

[@atharvaparulekar1035]

Its not as simple as that. Waymo is currently using LIDAR and HD maps, doesn't mean it will never move away from it. Using lidar data to segment cars, take decisions and train your vision networks can also be a way to approach the problem. I bet even tesla might be using some array of higher resolution sensors and high latency neural networks to annotate ground truth, segment objects on road for test vehicles etc. We know that self driving through vision is possible since humans only use vision. But is self driving with the help of vision the safest way to roll out self driving? Generalizing AI is not an easy task and self driving is an application where failure to take the right decision can cost lives. Waymo's approach is a safer approach on training wheels, while tesla's approach is harder to control but easier to scale. Waymo definitely will burn "way more" money than tesla in this process granted teslas path to level 5 autonomy is free of legislative, legal hurdles. But who will come out on top in the next 10 years is yet to be seen.

[@YR2050]

Tesla FSD beta can do all that Waymo does today, but Waymo can't do what a Tesla does, not by a long shot.

[@_____case]

@@YR2050 Waymo offers you a service in Phoenix (and soon San Francisco) that enables you to hail a driverless vehicle to take you from point A to point B with no one behind the wheel. Tesla has no such offering.

[@YR2050]

@@_____case Just because Tesla doesn't publicize doesn't mean they cannot do it. Tesla can easily navigate in low-density traffic without any intervention. The real difficulty is driving in downtown LA or NY. Basically, Tesla can do everything waymo can at the moment, Waymo cannot do what Tesla can.

[@_____case]

@@YR2050 San Francisco is not low-density traffic. I'm not saying Tesla isn't doing incredible things. I'm very impressed by what they've accomplished. But I'm also very impressed by what Waymo has accomplished. Another thing Waymo can do that Tesla doesn't do is ship goods between cities using Class 8 trucks.

[@skinnymoonbob]

2025: 11/17 TESLA just announced a partnership with STARBUCKS to deliver coffee at the doorsteps of customers. 2025: 11/23 WALMART announcement to partner with TESLA to deliver groceries within an hour to their customers.

[@saltyfish157]

lmaooooo crazy tesla fan boys are already starting to have delusion into the future.

[@allcouto]

Waymo is just a Train nedding tracks that does not change or it won't work. Tesla aproach in only vision deffinitively is the correct answer. AND Waymo is too much expensive to be a "comercial product".

[@5893MrWilson]

Exactly

[@charlesblithfield6182]

At about 630 it’s asserted humans solved the vision only driving problem. I disagree. Driving is a whole body kinaesthetic process and also involves sound perception. To what extent these make the rare case of the 99.9999% safe driver, or how much do the numerous accelerations on contact points of the body and the inner ear mean to good driving can’t say but driving is not vision only.

[@PunitSoni00]

It would be interesting (if possible) to have a human joystick a car with using only data that is available from Tesla cameras. If a human can do that safely, then tesla would have covered the minimum hardware requirements. If not, then their fundamental premise is wrong. After that, they need to see if they can build compute and software as good as human brain. Vision only driving is lot harder than Elon is making it to be.

[@charlesblithfield6182]

@@PunitSoni00 that is a cool idea.

[@pocnit]

Yeah and also humans are fucking terrible drivers so I'm not sure that's a great argument.

[@GreenisYellow]

Tesla will solve this sooner than expected data is key

[@neelsviljoen1383]

@@user-pz6mc2oj3f good data is key

[@5893MrWilson]

@@user-pz6mc2oj3f but Google doesn't really have actual driving data. You need more than a few dozen vehicles and a square mile in Phoenix Arizona. Tesla has billions of miles of data and the worlds most powerful super computer (dojo) solving this problem.

[@jessereeves3120]

That was the first variable they mentioned before rationalizing predictions

[@lemongavine]

@@user-pz6mc2oj3f data about what people search for online doesn’t apply to autonomous driving.

[@francescogiacomelli403]

Data and nn architecture, I believe Tesla is doing it so right (with the goal of automizing people’s transportation around the world)

[@davidapswoude3259]

Driverless value and market cap is hugely overstated . Paying double the going price for an electric versus ICE vehicle is not sustainable. Both companies are trading on pure emotion. Tick Tock-only a matter of time.

[@eanerickson8915]

Waymo has won. Elon is nuts.

[@MrGHawaii]

Dax Shepherd only became famous, the most forgettable famous person ever, because his parents bought him a spot in the movie Parenthood and other movies.

[@dc14522]

Tesla arguably has more data than it can effectively use. Project Dojo is a big part of the solution. Once Dojo starts to scale up I expect to see rapid improvements in Tesla's FSD. Interesting to hear that Waymo re-wrote the stack so it was more scalable. But what I want to know is how scalable are the creation and maintenance of the high resolution maps they need?

[@guslevy3506]

Every time I hear this debate these days, I feel like the Waymo methodology is to literally bootstrap street by street, neighborhood by neighborhood, city by city, etc. It essentially is attempting to solve a mathematical problem by churning and muscling through it step by step…and the only conclusion is that it will be linear growth model that will be complete only after centuries of work. Once one is through California, there is no factor which will allow it to learn the next areas in a quicker fashion. It is not an elegant solution…and for general autonomous driving, only an elegant solution will work.

[@bluetoad2668]

Fascinating watching the different approaches. Waymo have no choice, they don't have the data to do it Tesla's way. Tesla did have a choice, they chose vision only plus huge data. That's the key difference for me. Tesla had the choice of approach, Waymo did not. I don't know who will 'win'

[@ourv9603]

Currently WAYMO is on the road @ LEVEL4 while TESLA is still fumbling around @ LEVEL2. !

[@slick3996]

a year later from your comment and this is still accurate, meanwhile Mr.Musk has the nerve to addres it as level 5, hillarious

[@xploration1437]

@@slick3996 🤦

[@slick3996]

@@xploration1437 even funnier today 2 months later

[@xploration1437]

@@slick3996 Waymo is trash.

[@xploration1437]

@@slick3996 when did “Elon have the nerve to address it at level 5”? Now you’re just lying.

[@michellemenzies2187]

You should never wave a Pedestrian across, if someone else hits them it’s ur fault, it’s there decision to go.

[@calin2k]

this guy speaks like an apologetic MBA, not even answering questions.

[@vasumaru242]

I had a hard time understanding him. I feel like deep inside he wants to work for Tesla 😂.

[@cobrauf]

How will Waymo handle a broken traffic light with a cop giving hand gestures? There are a million edge cases like this. Lidar is a crutch, only Tesla has the right approach, tech stack, and data to solve this nearly impossible challenge.

[@PunitSoni00]

Why do you think engineers at Waymo have not thought about it?

[@pocnit]

@@PunitSoni00 Well I'm sure teams of the smartest people in the world working on this for years have never thought about a broken traffic light.

[@PunitSoni00]

@@pocnit /s ?

[@martindbp]

​@@PunitSoni00 Of course they've thought about it, but they have way less data for modeling those edge cases. Let's say you need to model plastic bags blowing across the road (as opposed to e.g. a ball). Whether you use lidar or cameras to determine whether an object can be safely driven over, you need to get that data. Tesla can ask their fleet of millions (soon tens of millions) of vehicles and get enough data back in a day. How long would it take Waymo? Multiply this by the long tail of edge cases and I think Tesla will reach L4 first, even with inferior hardware.

[@vasumaru242]

Waymo will get to L4 but not before Tesla. I agree with Tesla and Karpathy's approach to tackling autonomy. Data is huge value!

[@TheSpartan3669]

Waymo is L4

[@TheHeavenman88]

This guy didn’t say much in an engineering sense …

[@M.-.D]

Elon sets the bar too high.

[@supersonic956]

Is it even a competition? One of them is collecting real data from users using daily across the globe to imporve their software whereas the other one just has backing from Google for it to stand out as anything mildly significant.

[@mrneveryoumind]

Anyone still think waymo will win this 😬

[@ericmichel3857]

Seems like true L4-L5 autonomous vehicles will be just like the promise of fusion power, always just a few short years out of reach. Why do so many brilliant scientists and engineers always underestimate how difficult these problems really are? I have been hearing about the promise of fusion power for over 40 years, it is always just 10-15 years away! The most recent example was that lead scientist from Lockheed Martin(7 years ago) claiming they had it all figured out, he was talking about how he was going to need a new project soon because... Expected to have a prototype portable fusion generator in 2-3 years! Had it all figured out, just needed to solve a few engineering problems, piece of cake! That was seven-eight years ago, and now they have gone completely silent, imagine that. They also made a lot of bold promises about autonomous cars over the past 6-8 years, according to the hype we should of easily had L4 cars 2-3 years ago! Do scientists and engineers lie and exaggerate about the potential of technology? Say it ain't so! Sounds like we will have L4 vehicles around the same time our homes are powered by fusion generators on Mars. Anyone want to place bets on when that will be? If you are holding out for that robot girl friend, maybe think about a backup plan, just in case.

[@Ken-yp1dg]

Lots of problems could be solved by just lowering the speed limits.

[@Mizzle420420]

Essentially how to build a soldier that follows orders to the t with no human morality factor

[@dogmadan]

last!

[@0ooTheMAXXoo0]

Waymo aimed at level 4 since the beginning, Tesla aimed at level 5 from the beginning. Dude says things while avoiding the actual question asked. He is talking about some other unknown company when he talks about aiming at level 2... Long tail of computer vision is har he says. Waymo is trying to solve computer vision, lidar vision, radar vision, sonar vision and marrying all of them together. Tesla approach should be easier towards the long tail of problems...

[@minasithilindustries]

First!!!

[@michaelkartsonis6332]

Now I’m last

[@Kage1128]

Let's be honest Tesla already won, saying nobody won yet is the biggest COPIUM I have ever seen

[@Epiccurios]

"Waymo has an advantage" he obviously gets paid to say this 🤦🤦

[@bobboberson1928]

Nobody will solve this, it's a pipe dream like drone delivery.

[@eamonglavin2532]

I disagree it's certainly solvable (humans can do it) with a good enough ai it's obviously possible I do think that it won't be ready in the next decade though, city driving, unexpected events etc. Are extremely challenging to solve.

[@mrt094]

I'm starting to believe this more and more. I've been skeptical of these technological promises from companies unable to even put out a mass produced product. A lot of these tech companies just come off to me as over promising their way to the next round of funding.

[@mrt094]

@@eamonglavin2532 Eh my issue with that is whats really keeping them from reaching L4 now? It isn't hardware, It isn't money. Only thing really is software and the more I've been learning about A.I as of lately seems like A.I will never be what we want it be.

[@HsenagNarawseramap]

Tesla doesn’t need to get to level 5 to make money off it.

[@bobboberson1928]

@@HsenagNarawseramap yep, instead of deliving on anything they just make promises that things will happen, and they get more funding. But that will catch up when you don't deliver over time. Everybody is real happy with his internet service now.

[@mohamedlatreche9198]

Tesla is not even close

[@M.-.D]

What? Waymo is geolocked. It is like perfecting a Rubik cube solution and then calling yourself a problem solver……