Waymo CEO dismisses Tesla self-driving plan: “This is not how it works” • Hacker News Board

Re: Waymo CEO dismisses Tesla self-driving plan: “This is not how it works”

#451

Post by Fricken » Mon, Jan 25, 2021, 10:39 AM UTC

rcxdude wrote: post #411

mxschumacher wrote: post #73

Earlier quoted context omitted.

I should probably dig a little more into the microeconomics, but for me as a consumer the only difference between an autonomous car and an (human) Uber is price. If we assume that the price of the car stays the same, insurance and fuel still have to be paid and the cab app gets their cut - how much of my ride is really attributable to the cost of the human driver? So how much human labor can be automated away to the…

The driver is a huge part of the cost of a taxi ride (even with Uber subsidising them unsustainably with investor's money). An autonomous car company can price a journey at about the same as a bus ticket for a point-to-point on-demand journey and still make money hand over fist with their first generation hardware. Even for commuters this could be substantially cheaper than a car.

The running joke at Waymo for a long time was "How many engineers does it take to operate a self driving car?"

Re: Waymo CEO dismisses Tesla self-driving plan: “This is not how it works”

#452

Post by bartvk » Mon, Jan 25, 2021, 10:42 AM UTC

icebraining wrote: post #435

jjav wrote: post #424

Earlier quoted context omitted.

> You don't want to share a road with me when you could share it with a Tesla FSD. I'd rather share a road with you, a human. Even if you're a self-admitted bad driver, humans have a strong instinct of self preservation which helps. Software has no such thing, a bug in the code will let it accelerate full throttle into a wall (or whatever) without flinching because it's not alive.

Bugs in humans let them do that too: "The US NHTSA estimates 16,000 accidents per year in USA, when drivers intend to apply the brake but mistakenly apply the accelerator." https://en.wikipedia.org/wiki/Sudden_unintended_acceleration

Or: look-but-failed-to-see-errors, which are an "interesting" cause of accidents. When I took my motorcycle driver's test, my driving instructor sometimes warned me that I needed to make movements in a particular way. He claimed that even though I would make eye contact with a car driver, they may look-but-not-see-me. His reasoning was, as a motorcycle rider, I'm horizontal/upright when a car driver may be looking for something vertical (another car).

Re: Waymo CEO dismisses Tesla self-driving plan: “This is not how it works”

#453

Post by mikejb » Mon, Jan 25, 2021, 10:45 AM UTC

neillyons wrote: post #444

screye wrote: post #35

I worked with a self-driving product team a couple of years ago. Tesla's approach was looked upon with suspicion and often openly mocked. Tesla's entire self-driving strategy is led by Karpathy, who while brilliant, is an under-experienced researcher with a narrow expertise in large scale 2D CNNs. Tesla (and Comma.ai ?) seem to be the only groups that find Lidar unnecessary for L5 self-driving. I completely agree wit…

Elon talked about LIDAR at the autonomy investor day. Basically he said that the Dragon spacecraft has LIDAR and he spearheaded that effort. He also said that LIDAR and cameras both use the same wavelength, if you are going to add a sensor it makes more sense to add one which is outside of the visible wavelength. Even though I know nothing about driverless cars, the reasoning from Elon makes sense. https://www.youtub…

Whilst Lidar and cameras may have similar wavelengths, they have very different information outputs. Cameras give you a 2D colored image from which you essentially guesstimate the distance of objects. Lidar gives you a 2D depth map with much higher confidence in the data that comes out of the stream.

Elon has a high incentive to talk down Lidar in cars: He/Tesla placed the bet initially on cameras because Lidar is/was too expensive, and promised these cars will be able to drive themselves ("regulatory approval pending"). Retrofitting Lidar is not an option (price, sensor&cable placement), so a lot of his (and Tesla's) promises ride on cameras working out and Lidar not being necessary. A lot of Tesla's (Car's and Car company's) value rely on cameras working out and Lidar not being necessary.

Re: Waymo CEO dismisses Tesla self-driving plan: “This is not how it works”

#454

Post by theshrike79 » Mon, Jan 25, 2021, 10:47 AM UTC

_ea1k wrote: post #174

rainyMammoth wrote: post #146

Earlier quoted context omitted.

Sounds like lane steering and adaptative cruise control on my 19k$ Toyota Corolla.

How well does that handle poorly marked sections of road? I have a similar system on a non-Tesla and that was one of its main weaknesses relative to AP.

Usually they need clear lane markings, at least one white line needs to be visible for the system to latch on.

Snowy roads, dirt roads etc. are out of scope for those.

Re: Waymo CEO dismisses Tesla self-driving plan: “This is not how it works”

#455

Post by theshrike79 » Mon, Jan 25, 2021, 10:49 AM UTC

burmanm wrote: post #277

Geee wrote: post #259

FSD is all about collecting insane amount of data about edge cases that happen in the real world. Tesla has built the infrastructure for this, and will win the FSD race by a wide margin.

Is snow an edge case? My cameras don't function when it's snowing or there's snow in the ground. Or there's snow banks around the road. My dash is lit with the "some cameras are blocked" 24/7 for the whole winter time. And proximity readings are blinking those yellow / red lights all the time from the right side as it thinks the snow bank next to the road is too close. Or the system thinks the camera is blocked becau…

Snow is a difficult problem even for humans.

In worst cases the road turns from a 2 lane road to a 1.5-1.8 lane road when there's enough snow on the banks. Multi-lane motorways turn from 4 lanes to 2-8-3.5 lanes. Humans have trouble navigating that, dunno how any FSD system would manage it in our lifetime.

Re: Waymo CEO dismisses Tesla self-driving plan: “This is not how it works”

#456

Post by Erlich_Bachman » Mon, Jan 25, 2021, 10:52 AM UTC

jfim wrote: post #330

nickik wrote: post #208

Earlier quoted context omitted.

Nobody in the autonomous car industry had produced anything yet. Tesla has the best system of those actually in use. Their arrogance is about as convincing as that of the rocket companies or car companies that didn't think EV were gone happen. Saying 'in my industry they are joke' is kind of a joke by itself by now. Also Tesla wants to operate a taxi services too so I don't understand your argument.

> Nobody in the autonomous car industry had produced anything yet. Tesla has the best system of those actually in use. Waymo is actually operating a driverless paid service in Arizona [0] with point to point pick up and drop off. Does Tesla actually offer this service today? [0] https://www.wbur.org/hereandnow/2021/01/04/waymo-driverless-...

> driverless paid service

Which works in one small limited area of the earth because of high-def maps, which is not what you would normally call "scalable".

Re: Waymo CEO dismisses Tesla self-driving plan: “This is not how it works”

#457

Post by Erlich_Bachman » Mon, Jan 25, 2021, 10:55 AM UTC

YZF wrote: post #170

onion2k wrote: post #57

Earlier quoted context omitted.

Tesla (and Comma.ai ?) seem to be the only groups that find Lidar unnecessary for L5 self-driving. There is another group who find Lidar unnecessary for driving: humans.

I own a model 3. Often it says "the camera is blinded" (e.g. when the sun is shining at it), often there's rain on the lens... So I'd argue Tesla's cameras are much worse for driving under all conditions vs. the human eye. I'm not sure there's enough redundancy for depth perception either. I tend to agree with the argument that humans prove it's possible to drive with just visual sensors but Tesla's sensors, for bett…

> Often it says "the camera is blinded"

It could very well theoretically be a software issuse.

Re: Waymo CEO dismisses Tesla self-driving plan: “This is not how it works”

#458

Post by Erlich_Bachman » Mon, Jan 25, 2021, 10:58 AM UTC

sudosysgen wrote: post #356

jfoster wrote: post #297

Earlier quoted context omitted.

It's fairly easy to dismiss that argument; the cameras would not be difficult to replace if they are the limiting factor. Cameras are cheap.

The human eye has capabilites that you'd need a camera more expensive than a LiDAR by a good bit too replicate. No rolling shutter, very low delay, continuous signal processing instead of per-frame, maximum resolution around ~70MP eq. in the center, f/2.8 aperture at a full-frame image size, servo driven active rangefinding, continuous cleaning, etc... A camera with those features would be around 3000-15000$, and you…

> Also, the human brain can use focus information and stereo phase to deduce 3D structur

Current AI tech can do this too, from multiple images. Karpathy even talked about how they do use that in Teslas.

Re: Waymo CEO dismisses Tesla self-driving plan: “This is not how it works”

#459

Post by mikejb » Mon, Jan 25, 2021, 10:59 AM UTC

kerng wrote: post #311

newbie789 wrote: post #262

Earlier quoted context omitted.

That’s a very interesting neutral take that “Tesla’s strategy might be the winning horse in this race” and that the Waymo CEO may be defensive or scared. I personally don’t know enough about either company or their tech or how people react to self driving cars to call a winner when it comes to a something (in this case, fully self driving cars at scale) that literally don’t exist yet with enough certainty to predict…

How else would you describe Waymo's message? Trying to "educate" or talk down to a competitor with words like "this is not how this works" when Tesla has significant resources, expertise (actual cars on the road!) and knowledge in this area seems defensive to me.

To be fair, Waymo (back then Chauffeur) have solid experience with Tesla's approach. They started with the same strategy of gradual improvements. They found out (and shard) the same issues Tesla is running into in the past years: Users don't pay attention when monitoring a system doing a boring task - even when failures carry a high price. Waymo switched the approach from a gradual L2/3/4 improvements straight to developing an L4 system; Tesla is locked in to the gradual move due to it being part of their business model, and are now stuck with trying to make self-driving work with unqualified test drivers (i.e. customers) and outdated technical decisions (not using Lidar because they were too expensive back then).

So Waymo does have the experience to call out Tesla here.

Re: Waymo CEO dismisses Tesla self-driving plan: “This is not how it works”

#460

Post by mikejb » Mon, Jan 25, 2021, 11:00 AM UTC

cbozeman wrote: post #282

stingraycharles wrote: post #268

Earlier quoted context omitted.

They are collecting data on all cars yes, but are training a model in one place. That is not federated machine learning, it’s federated data collection at best.

I think the massive collection of data from more and more Teslas is going to eventually allow them to produce the best self-driving vehicle. Its been a maxim of history that those with the most data generally win, whether that's military intelligence of troop movements from 2000 years ago, or LIDAR scans from a car today.

This is a curious point: How much data does Tesla actually collect? I can imagine users not being too happy if their car clogs the upload for hours every time you park it in your garage?