Hacker News Reader

> Google's LIDAR technology is obsolete spinning-scanner gear, mostly from Velodyne. It's something you'd use on an experimental vehicle, not a production one.

Given that (1) Google is spending a whole lot of money to protect their LIDAR technology in court, (2) Uber spent nearly half a billion dollars apparently to copy it and (3) Google (with their proprietary LIDAR) is the only company -- publicity and press releases aside -- that seems to be actively operating an autonomous fleet in the wild with reasonably few overrides, disengagements or accidents, I'm willing to bet that good LIDAR is actually more important than you're giving it credit for here.

So a good LIDAR steers a car all by itself?

I'm willing to bet that there is a ton more of technology involved in self driving vehicles than you are giving it credit for here.

Not sure why the down votes. There is a lot more technology here than one type of sensor, there ia a huge software stack. Uber clearly paid $500M because they wanted to get access to someone who had been doing it for a decade and had built that stack more than once. Whether Uber thought they were getting Google technology in the deal is the big question of the lawsuit, but it's very possible they just wanted to jumpstart their own effort by buying a team that was already up to speed.

And the other question is why would auto makers want to sell or sacrifice their production capacity so they Google can kill them?

Also, there are soy many regulations and safety rules in the industry, if they can't abolish them (with perfect tech/lobbying) old automakers will win.

> And the other question is why would auto makers want to sell or sacrifice their production capacity so they Google can kill them?

Good point, but I think there are many ways to make this work. Two ideas:

1. Contract with a small/failing auto manufacturer who has little to lose (e.g. Mitsubishi).

2. Work with a manufacturer to build vehicles and also license them tech (Chrysler and Honda are interested [0]).

[0]: https://www.forbes.com/sites/alanohnsman/2017/01/10/alphabet...

If you bought a new Ferrari because you were tired of the old one, and someone steals your old Ferrari do you just let it go, not charge them with theft?

Maybe Google thinks they can drive a wooden stake through the vampire heart of Uber with that LIDAR claim. Perhaps they will if this turns out to be a valid claim.

>Google's LIDAR technology is obsolete spinning-scanner gear, mostly from Velodyne. It's something you'd use on an experimental vehicle, not a production one. The production LIDAR systems are coming, they're all solid state, and they come from big auto parts makers like Delphi and Continental. So by the time the Uber case gets to trial, it will be moot.

Google's LIDAR technology is obsolete? Didn't Google just come up with a new system that reduces the cost of a LIDAR system by 90% ?

https://www.bloomberg.com/news/articles/2017-01-08/alphabet-...

The 90% cost reduction from the original $70,000 Velodyne spinning KFC bucket is still nearly 2 orders of magnitude more expensive than the expected prices for solid state lidar, which a whole bunch of companies including Velodyne are racing to develop for the mass market, which are expected to be as little as $80, though several units will be needed to provide complete coverage around a car.

>which are expected to be as little as $80

I've been paying close attention, and the lowest price I've seen has been $300, and I don't believe any solid state lidar has yet to be produced at that price point, despite companies claiming they could more than a year ago.

'expected' is the key word here. Getting an esoteric new technology from a proof of concept in the lab to the mass market doesn't happen overnight, but when it does, and after it's been through a few product cycles, it tends to get cheap.

Right, technology advances over time. Not sure what you're arguing here. Nothing is on the market at a lower price point, and the conversation you jumped into was in response to someone calling LIDAR obsolete. Of course something will eventually come out which is better and cheaper, but it ain't here yet.

Velodyne has announced they'll have early production models for testing this summer, with full production before the end of the year. The lawsuit won't even be done before the technology they're fighting over is obsolete. C'mon, now. It just shows the lawsuit isn't actually about protecting some kind of invaluable trade secret, that's just a tool for punishing Levandowski and Uber for hatching this scheme.

That's an awful lot of assumptions in one sentence.

Flash LIDAR from ASC [1] works fine. DoD and Space-X have been using it for years. It costs $50K/unit because it has custom ICs made with a nonstandard InGaAs process produced in tiny quantities. The units are hand-made in Santa Barbara, CA.

Continental, the big German auto parts company, bought ASC, and is putting the technology into mass production. Continental says they can get the price way down, and they probably can.

There are also about a half dozen other flash LIDAR startups, some of which might even ship someday.

[1] http://www.advancedscientificconcepts.com/products/peregrine...

Everything you described in the last paragraph is a commodity market other than software. The margins will be in software or the overall car brands themselves (since vehicle manufacturers are large enough to throw their weight into lowering software margins assuming there's more than one option e.g. not like Microsoft in the 90s).

Levandowski has been invoking the Fifth Amendment in this case, which is a case that he's not even a party to. That's either legal gamesmanship, or else he has a genuine fear of prison time out of this. (Alternately, it might be possible that he could lose a big chunk of money but avoid prison.)

If he comes out of this still holding a big chunk of money, but having to spend several years in prison, is that still "coming out of this really well"? In my view, no.

Didn't the judge order last week that Levandowski couldn't invoke the Fifth Amendment for that very reason? Or rather that it couldn't protect him?

Source: http://www.businessinsider.com/uber-exec-anthony-levandowski...

Yes. If Levandowski is not just playing games, that decision is not going to decrease his fear of prison...

Not a lawyer, but it's my understanding there is no 5th amendment protections in a civil suit/trial. But Googles complaint frames Levandowski's behavior as criminal, so doesn't that create a reasonable fear of prosecution, and open the door to letting Levandowski use the 5th?

I don't know. But it's an interesting case.

> Not a lawyer, but it's my understanding there is no 5th amendment protections in a civil suit/trial.

It was weirder than that. Levandowski was claiming that Uber couldn't turn information over to Google because it might incriminate Levandowski, when Levandowski wasn't one of the parties in the lawsuit. Unsurprisingly, that didn't fly.

You can assert the 5th amendment in a civil suit, if you fear potential criminal prosecution. If you couldn't, then the 5th amendment protections would essentially disappear in many cases. Many criminal offenses also give rise to civil liability. So, all that would have to happen is for the victim to sue the criminal defendant, notice his/her deposition in the civil case, and compel the criminal defendant to speak that way. Then you would suddenly have the defendant being forced to testify under oath, despite the fear of criminal prosecution.

So, Levandowski was trying to use the 5th Amendment to prevent Uber from disclosing, during discovery, the author of a report (in Uber's possession, not Levandowski) that Uber was claiming was privileged.

There are a few issues here: 1.) 5A protections don't generally cover the production of existing documents, only testimony (statements, answering questions). 2.) 5A doesn't protect corporations, only individuals. 3.) It's (apparently, per the discussions in the various court filings on this issue) incredibly rare to try to prevent the disclosure of the metadata on a privilege log. (Basically, during discovery, if you find a document that's responsive to the discovery, but think it's privilege from being turned over, you have to provide a report of what the document is, who has it, who has ever seen it, etc. to prove that it remains privileged.). 4.) For someone who isn't a party to the lawsuit to try to assert this is incredibly unusual.

The judge basically called the whole argument bullshit. But it was all related to the document stuff, not Levandowski's testimony/deposition.

> Levandowski has been invoking the Fifth Amendment in this case, which is a case that he's not even a party to.

Levandowski is an intervenor defendant in this case.

The disruptiveness is measured by the effect, not the effort.

I believe Animats was referring specifically to disruption of industry incumbents and failed to clarify that.

The disruption is mostly social. We will need less parking spaces, driver's licenses will become increasingly unneeded, and car ownership will largely become something for hobbyists. Transportation will become even more a subscription, with the only exceptions air and sea travel.

Waymo's endgame here is intellectual property licensing. They're not trying to be in the business of manufacturing sensors and actuators. They're trying to be in the business of licensing software that makes a working self-driving system out of a car equipped with sensors and actuators.

"Having our hardware and software development under one roof is incredibly valuable. Our sensors are developed in close collaboration with our software experts who specialize in AI techniques like machine learning. All of our sensors — including LiDARs, vision system, and radars — are deeply integrated with the brain of our self-driving cars and specially designed for our software. A single integrated system means that all the different parts of our self-driving technology work together seamlessly. Like a person’s own five senses, our sensors are more useful and more powerful when we put them all together."

https://medium.com/waymo/introducing-waymos-suite-of-custom-...

It's interesting that all the self driving software seems to be advancing at a similar rate - Google's, Volvo's, Tesla's and Mercedes seem to have some what similar performance, which I guess is because most AI research is published and people use similar techniques. At that rate there won't be one system dominating but a bunch of comparable ones probably including third party suppliers that any auto manufacturer can bolt on.

Apple is an example of why your last paragraph likely isn't true. It's about the end-to-end experience, not the parts that compose it.

I don't know about you, but for the first couple of years at least, my primary concern when buying a self-driving car would be making sure it's advanced enough to not kill me if it hits an ice patch, not the "fluid end-to-end user experience" it offers.

Exactly. The quality and craftmanship of the product is what should be priority (especially concerning people's very lives), the end to end should come naturally for any respectable company.

I agree. I was responding to someone who thinks that will be commoditized by "auto parts companies". Whether or not that's true, safety is the most important component of the user experience of a self-driving car.

Once that's figured out, the question turns to "Am I buying the car to own? Am I calling it from a mapping application like Uber or Apple Maps? What does the car look like? What new seating arrangements are enabled without the need for a driver? How much better can the entertainment features be when built by a software company instead of a car company? Can the car tell me what's wrong with it, instead of simply illuminating the check engine light? Or better yet, automatically contact the manufacturer to set up a service appointment that works with my calendar?" Etc.

> Levandowski already got his money.

Did he get straight cash, illiquid uber stock or did he get a big contract with a lot of performance stipulations in it?

Might be a reference to the fact that Levandowski already made $100M+ selling his first company to Google. I would bet that his team got almost entirely Uber stock in the Otto acquisition and that there are vesting provisions. My reasoning?

1) It would be silly for Uber to blow through their cash that fast and 2) silly to give a new team that had only been together for a month or two a huge pile of cash they could immediately walk away with.