I worked at 510 for 2 years. We were a little over 10 people when I started and around 60 people when we sold to Google. The whole thing was bootstrapped, no VC's. Bryon Majusiak, the guy interviewed in the article, is a great guy, a hard charger and a very practical engineer. So much so that I hired him away from Google a little less than a year ago for another robotics endeavor.
510 did many great things and got credit for nearly none of it. We barely had a website b/c big companies like Google, Microsoft and Nokia liked letting people think it was all them. Quite frankly, we were happy to take the money, but in hindsight, we should have aimed higher.
The article gives the impression that we only did hardware. That is not true. We did all the sensor integration and real-time systems software. We also sold our GPS filter to Google early on for Street View.
It was a scrappy place. We solved hard problems very quickly and big companies often resented us for it and reluctantly bought our wares. In return we were silent.
The employees were fantastic and the management was horrendous. We laughed a lot, we fought a lot and we drank a lot. The founders, like most 20 somethings who've achieved too much success too early, were amazing in many ways but deeply flawed. They did not realize that the Berkeley Robotics Lab did not prepare them to manage a real company. They could never see that they were the ones in the way of their own success. Since there were no VC's there was no board. I never again worked for a place where the CEO could not be fired. And I never worked for a founder who had too many life lessons in front of him.
Anyways, lessons learned: The value is in final products not in parts of products, people don't change on a startup's timeline and getting acquired is generally a letdown.
Well, Lewandowsky heads Google's self driving car project now. Thrun is long gone, off trying to make Udacity fly.
He's an impressive guy. I met him when he was going to Berkeley and doing the self-balancing motorcycle for the 2005 Grand Challenge. He'd already done a startup, with a specialized giant laptop for construction sites for people who needed to see blueprints.
The LIDAR Google uses is from Velodyne, which was "Team DAD" in the 2005 Grand Challenge. The first version of their LIDAR fell off their vehicle, but they improved the mechanics and produced that cone-shaped thing Google now uses. That's really a research tool; a different approach is needed for production vehicles. (I still like the Advanced Scientific Concepts flash LIDAR; it's expensive, but that's because it has custom silicon. If you had to get the price down, that's where to start. No moving parts, all electronics.)
I'm kind of disappointed with Google's self-driving effort on the hardware side. I'd expected flash LIDARs, terahertz phased array radars, and other advanced sensors by now. You need to be able to see in all directions, but the requirements to the sides and back are less than for looking ahead. The CMU/Cadillac effort is ahead on the hardware side; their self-driving car has all its sensors integrated into the vehicle so you don't notice them.
(I had an entry in the 2005 Grand Challenge: Team Overbot. Ours was too slow, and we worried about off-road capability too much.)
I suspect price is still the driver (pun intended) here. Flash LIDAR is, for the most part, shockingly expensive. Some companies like Toyota are looking into bringing the cost down with CMOS SPAD arrays, but it's still very much lab work. Similarly the Velodyne system ain't cheap and was close to $80k not too long ago. Stereo is pretty much free, but still needs more work to make it robust enough. Even smaller line scan LIDAR units (e.g. the SICK LMS models) are around $5k each.
I recently learned that Elon Musk wasn't actually the founder of Tesla, but rather an early investor. This article borders on the same theme; there's more than meets the eye to the faces of today's most innovative technologies. It is alarming to see how often proper accreditation is being misattributed/stolen in our industry.
Generally, I agree with your sentiment regard proper accreditation in the media, but claiming that Thrun doesn't deserve a huge chunk of credit for self-driving cars at Google is thoroughly unsupported by the evidence (check out the history of the DARPA challenges). This article seems to be grasping at a sinister story of a coverup that simply doesn't exist.
From the article: "From then on, we started doing a lot of work with Google," says Majusiak. "We did almost all of their hardware integration. They were just doing software."
Just software? Holy cow. The machine learning, control systems, mapping software, and all the other algorithms in a self-driving car are the really fundamental pieces. It's super important to get the hardware right, and 510 did a great job of that, but the software is the brain.
Technically, I don't think Musk was the founder of Paypal either. I think he founded x.com, which merged with Confinity (the company founded by Max Levchin and Peter Thiel - who had started PayPal).
I was quite surprised by the parent comment because I had never heard of Musk being described as a founder. But then again I've lived in palo alto since before tesla was founded so the power struggle was minor gossip at the time (and was in the mercury news).
What has surprised me is that he is in fact listed as a founder on the wikipedia web site and in fact Tesla litigated over the matter.
Now the word "founder" is weird -- IME A round paperwork typically refers to any common shareholder at the time of venture investment as a "founder". And then there was that bizarre Facebook suit over who could refer to themselves as a company founder. Weird.
"Founder" title drama is mostly around successful companies in my experience. Basically some additional mojo can be had a funding discussions if you can legitimately claim to be a 'founder' of some previous success (the bigger the better). Especially in social settings it is strange (and to my twisted humor funny) when people will say they 'worked' at one company and 'founded' another company when they had the exact same role in each company at the same level of development and differ only in the perception of 'success' or 'not success'.
After reading the article on how primates use 'fame' in dominance games that made the rounds here I found that it identified that sort of behavior pretty precisely.
I think the bigger point is that Google has created a revisionist story about the origin of its self-driving car that the vast majority of people, including myself, would never have known about. For example, there was no mention of it at this summer's self-driving car exhibit at the Computer History Museum in Mountain View. Is it necessary that Google represent the true story in its marketing material? No, of course not, but it is strange that they would go to that extent to keep the acquisition a secret for so long.
Have not seen the museum. So long as they are crediting the engineers from the DARPA contests then the story is not revisionist. Everything since DARPA has been fairly incremental with contributions from a wide variety of people (including 510). There just isn't a Da Vinci code type conspiracy going on here.
The article presents a one-sided story at best. I actually interviewed at Google for "project that shall not be named" in June 2009, with Sebastian Thrun and Chris Urmson (the current head of the project). I could tell the project was already well underway. Although none of them would say anything about, it was pretty clear what they were building. So, while '510' may have played some part, the writer is grossly exaggerating their significance.
Maybe I'm missing something, but isn't the lidar on the google car made by Velodyne?
I know the radio head video used Velodyne's Lidar.
And the picture of the lidar that they show is a picture of the Velodyne HDL-64
Can anyone explain?
The article didn't really make it sound like 510 built the LIDAR, just that they do the integration and software to process the data into maps. The article is also very explicit about their role in processing the data into the cool cloud visualization for the Radiohead video, not capturing it.
I think the photo captioned "510’s lidar system" and showing nothing besides than a Velodyne sensor bolted to a frame is specific enough to be misleading.
Velodyne already has their own self driving car. What google is working on is the hard bits like figuring out how to handle cyclists, shitty drivers, pedestrians, traffic lights. Those things are hard.
But someday I'd really like to see an open source version of a road capsule.
Mainly aimed taking existing cars and running them like trains on a track.
1) Stop when an obstacle is in front
2) Automatic follow the current lane, maintain a safe distance with car in-front and follow the speed limit.
3) Parse traffic lights and be able to make turns.
4) Wait for passenger to get in and out.
There's no reason why I can't take a short nap while I'm on a 100km straight road going cross country at night when there are almost no cars on the road.
I'm betting that someday in the future we're going to have lanes on road specifically for self driving buses/cars just like we have dedicated train lanes.
One of the lessons of the Grand Challenge was, put a bunch of sensors up there, even though it's expensive, and be sure you have good lidar coverage. Velodynes are one main high-end option for lidar.
Lidars can be more effective than stereo cameras (although they have tradeoffs vs. stereo in some use cases).
“From then on, we started doing a lot of work with Google,” says Majusiak. “We did almost all of their hardware integration. They were just doing software. We’d get the cars and develop the controllers, and they’d take it from there.”
Anybody who has done robotics knows there is a lot of integration involved (hardware and software), and that doing integration is hard and tends to be thankless. It's nice to see some in-depth reporting in a major publication on the full depth of the engineering team.
Good question! I'm not going to post details here, but she appears to have plenty of public info easily Google-able, including a Linkedin profile and the name of her current startup. That said, I couldn't find anything about this company at all, and don't want to delve too much deeper as it's starting to make me feel creepy.
That's because you're not. You posted a couple of comments from an IP that was banned because of past abuses (apparently by others). We unbanned the UP and unkilled the two comments.
It's on our list to make the IP banning system better so that this happens less often. Having it happen at all is bad, obviously, but the measures available to counter abuse are often unsatisfactorily crude, and doing nothing is not an option.
My apologies, I should've just said "banned" or "dead" or some other word. I've seen other people use the "hellbanned" word before and figured it was the right phrase.
Out of interest, I did look up the user and saw that he/she seemed not to have done anything that I felt deserving of any form of ban - how should I have communicated this back to you? Am I even encouraged to do so?
like with Internet, the self-driving cars revolution is firmly rooted in DARPA's Grand/Urban Challenges of 2004/2005/2007 years. In fact i have hard time finding any principal progress in Google cars from the winners/top cars of 7 years ago. Though obviously there is a lot of evolutionary/product-development-style improvement, especially related to increased processing power available for sensor/image data processing.
That tends to be the case with all technologies though. Initial progress at adding features is much greater (to an outside observer) than later progress. What you probably don't see is the large number of small incremental improvements that are slowly ironing out edge cases, fixing bugs and turning prototypes into something reproducable. I bet if you ask any team involved in those challenges you'd find that the apparently functional cars were constantly requiring fixing and tweaking in a way that the Google cars of today probably don't.
Agreed. There was an immense amount of progress between 2004 and 2005, and then another big leap for the urban challenge. I was expecting them all over by now, but those prizes really motivated people. I've personally put autonomous vehicles on the back burner in favor of other projects because without that juicy moneybag it just seems like too much of a long-shot to work on them.
Someone really needs to explain to me what the thought process behind this acquisition was.
Self driving car technologies have has actively developed by almost every car company for years now. Many of the beginnings of this work has already made it to market e.g. Parallel Park Assist, Auto Emergency Breaking, Lane Merge Detection, Adaptive Cruise Control. And companies like Volvo are already testing their self driving cars in real world, difficult conditions in Sweden. And because there are only a few car conglomerates they will simply share technology within each group.
1) To supply the data that the future robot cars will consume, and profit via license fees. The Google cars use data-heavy techniques with very good preexisting metric maps.
2) To speed up the development of technology that will save perhaps 20,000 lives a year in North America alone. Larry Page has enough power and Google enough money that they can do stuff with "because it's great" as a primary reason.
(I'm not affiliated with Google, but have pondered this question and asked around)
Volvo is actually far behind -- their cars are designed for Level 3 Autonomous Driving, while Google cars are Level 4. Also, Google has far more miles in real-world city driving than Volvo.
Right now Google is furthest ahead in this space, arguably by years.
Where have you heard this ? Everything I've seen to date has Google/Volvo at the same level. And it's somewhat irrelevant how many years Google is ahead when they have zero experience in bringing automative to market.
510 did many great things and got credit for nearly none of it. We barely had a website b/c big companies like Google, Microsoft and Nokia liked letting people think it was all them. Quite frankly, we were happy to take the money, but in hindsight, we should have aimed higher.
The article gives the impression that we only did hardware. That is not true. We did all the sensor integration and real-time systems software. We also sold our GPS filter to Google early on for Street View.
It was a scrappy place. We solved hard problems very quickly and big companies often resented us for it and reluctantly bought our wares. In return we were silent.
The employees were fantastic and the management was horrendous. We laughed a lot, we fought a lot and we drank a lot. The founders, like most 20 somethings who've achieved too much success too early, were amazing in many ways but deeply flawed. They did not realize that the Berkeley Robotics Lab did not prepare them to manage a real company. They could never see that they were the ones in the way of their own success. Since there were no VC's there was no board. I never again worked for a place where the CEO could not be fired. And I never worked for a founder who had too many life lessons in front of him.
Anyways, lessons learned: The value is in final products not in parts of products, people don't change on a startup's timeline and getting acquired is generally a letdown.
Onwards and upwards...