In my previous post I shamelessly bragged about my recent graduation from the superb Udacity Advanced Self-Driving Car Engineer Nanodegree Program. That answered the question, what has been keeping me so busy this year? Now I will address part two, why. Why would someone do this and specifically why would I?
Let’s start with an obvious standard question. Would I recommend this educational experience to others? My answer is perhaps surprising—not necessarily. Make no mistake, obviously Sebastian Thrun knows probabilistic robots. Course director David Silver is a super cool guy and a terrific guide to this challenging material. All the instructors were excellent. Everyone I interacted with, both actively and passively, earned my respect. The course material was extremely solid and well organized. I felt that the Udacity platform and support were all well in line with the tuition expense.
Furthermore, and this is important, I feel like if I had to hire engineers for my autonomous vehicle project I would heavily favor graduates of this program. I know for certain that I did learn a huge amount about the topic. This was absolutely a very effective use of my time and resources.
So what could possibly be problematic? Well, on the SDCarND homepage Sebastian Thrun (president of Udacity, famous AV pioneer, lecturer in the course) says this.
There’s an enormous market for self-driving car engineers. Lots and lots of companies that you wouldn’t suspect have entered that field and are massively hiring.
When I first started the program this kind of credulous promotion seemed harmless enough. It can’t be denied that in the last year there has been a bubbly rush to join the autonomous car fad. GM and Ford have each spent a billion dollars on autonomous dreams. There are a lot of start ups getting tons of money. And we can plainly see from the news that Tesla and Uber and Waymo/Google all are involved. And no matter what the level of investment today, intelligent people clearly see that the technology exists to make our current transportation modes obsolete; the only question is how exactly will that happen and when? Given all of this, it may seem reasonable to assume "massive hiring".
But there’s one area where I am far more proficient than Sebastian Thrun and that is simply being a normal guy. I excel at not being a celebrated professor of robotics from one of the world’s top technical universities. I don’t like to brag, but I’m even pretty good at not having a PhD. So when Professor Thrun says there is "massive" hiring, I feel like he means that everywhere he goes, everyone he talks to says the same thing, "We just can’t hire enough qualified people." And that’s probably what Thrun hears, but what is actually meant is "We’re having trouble hiring an army of Sebastian Thruns for a business unit that generates no income."
One of the wonderfully valuable aspects of this course was interacting with other students in the forums. I want to highlight this observation someone posted.
Does anyone want to share their experiences from the hiring event the other night? I had a great time there, but for me it wasn’t all positive. For example, the recruiter from Lyft basically cut me off and said "We are looking for people with industry experience." And then I tried to go in depth about some of my projects from the nanodegree, including the Bosch path planning challenge, and she cut me off again and said something like "Unfortunately the Udacity path planning material is not enough for us."
It seemed to me that strong programming skills, specifically in C++, were more important than the concepts covered in this course. The person that I talked to at Uber told me that they pretty much only hire people that have at least a masters in computer science.
I feel this is representative of what I heard from many people (in my final project team, for example) and what I even experienced myself with two major players. This shows a couple of dissonant ideas in the autonomous vehicle business. The first is looking for people with "industry experience". Obtaining exactly that was definitely a minor reason I embarked on this course. And I feel I got it as good as it gets. But the more glaring point is this: there is no industry experience! This industry does not exist! Obviously decades in the computer business show that a couple of years after a fad technology arises, all businesses will be found trying to hire anodyne salarymen with "5 years experience". So par for the course there.
But the second point this shows is that precisely because this has not evolved into a positive-cash-flow "industry" it does not need "engineers". It needs researchers. It’s little wonder that these companies are gutting top university robotics research labs.
My own personal experience has not been much different. Would you rather have some random PhD who needs a job or some insane guy who has made it his personal crusade/hobby to tirelessly accomplish your company’s goals? What about two of the latter (they’re cheap) for one of the former? Nope, that’s just not a thing in this putative "industry".
I do get the feeling that Udacity has taken great pains to try and facilitate finding a job. They’re doing everything they can I’m sure. I think it spills over from their other offerings where there are (unfortunately) millions of Android and web developer jobs, etc.
According to this article the Udacity Self-Driving Car Engineer Nanodegree program has accepted 10,000 students. Shockingly, 43,000 have applied! I know that at the vanguard of this the quality may be a little higher, but the people I observed and interacted with were extremely high quality! If there’s one thing I would observe about this "industry" it is that there is no shortage of extremely competent people who are extremely interested in doing this work. These are great people, serious people. If they are not getting hired by companies that believe they need self-driving car engineers then those companies are deeply flawed.
During the final term my favorite tier one auto supplier, Bosch, sponsored a contest involving autonomous vehicle path planning. The top 25 participants would win an interview for a "Research Engineer" position (note the self-contradictory job title, is it "engineering" or "research"?). Without going into the bizarre miracle that allowed me to be one of those top 25, I did get to talk to the person who makes hiring decisions for Bosch’s autonomous research group. It turns out that while Bosch is a big company with 390,000 people, their Palo Alto offices have only 200. And unless I misheard, the autonomous vehicle group there has only twenty-five people. And as I’ve pointed out before, Bosch is one of the major players in this "industry". If they can’t easily double or quadruple their tiny AV gang from the 43,000 people who are keen to help them do so, I’m sorry, but something is very wrong with that picture.
Then there is the problem of the potential workforce not being Sebastian Thrun enough. God knows Professor Thrun has done everything anyone can be expected to do to share his expertise! What these foolish companies do not understand is that the people who have just completed this course are not cast offs or rejects from "proper" educational attainment. These are very high quality people. They have shrewdly optimized their educational investments. So much so that I will assert that you won’t find anybody better. The reason is that if you are clearly a cut above the people I saw in the program, you can probably get NSF (and other) grants and basically research whatever strikes your fancy on your own terms. Or, quite realistically, you can start your own business.
Working for a big well-funded company can certainly have advantages and I’d love to work for a group with their own private test city to play in. But that’s not infinitely valuable or even strictly necessary at this stage. There is still a lot of software to be written and that costs exactly nothing to get started with.
Standing on the job seeking sidelines it’s been a good show. I’ve played along with the "career development" stuff Udacity offers, but I didn’t take this course to "get a job" (got one thanks). I took this course because I literally can not wait for this problem to be solved by others. I genuinely wanted to know as much as I can know about how best to reduce the amount of human stupidity involved in driving. If you share that goal, then you will be delighted with Udacity’s course. If you simply need a job, well, I’m not so sure it will work out. (I’m totally open to being corrected here but I’m just calling it like I see it.)
When I look at the flurry of recent interest in autonomous driving I am reminded of bicycle racing. I know most people think it’s a very boring group of guys riding bicycles from point A to point B, but it’s actually one of the most interesting game theoretic sports there is. In cycling, riding quickly is actually the anti-goal. The goal is to do as little work as possible on the way and yet still win at the end. I often feel like the major players in the auto industry are mostly marking each other rather than giving it their all; in fact, that is their correct strategy! Very often some antsy rider (stupidly) can’t hold back his wattage and breaks away from the comfort of the main pack’s slipstream. These can often lead to very long punishing solo ordeals. The pack makes sure that they never let the breakaways get too far ahead and, indeed, most are dramatically but precisely caught just a few kms from the finish. Sometimes, however, the breakaway artist pulls it off, the powerful favorites in the main peloton misjudge his strength (or their own laziness) and they can’t quite catch him in time.
My Udacity experience was just one phase in my solo attack. I do not expect to win. I’ll be disgusted if I do! That is not my goal. But know this — a race with a lot of spirited attacking always finishes faster than one without challenges. And I am in a hurry to see the finish line where the misdeeds of distracted, drunk, and idiot drivers are merely a chilling historical footnote rather than something I rationally fear for my life about every day.
I hope that SDCarND graduates do not wait for a job if it is slow to come. Attack this now. Do what you can to move autonomous vehicle technology forward. A fool like me will surely fail, but 10,000 of us fools will not.
