Apple Putting the Brakes on its Self-Driving Unit is no Surprise.
The decision has been years in the making and a reflection of an industry that is trying hard to counter-steer.
It’s 2017 and the air buzzes with the promise of an autonomous mobility revolution. I had just hopped on a plane to pursue a master’s in AI with no clear idea of my intended specialization. Fast forward six months and every needle pointed towards a glorious future in self-driving. My graduate advisor shared with me a list of potential career pathways. All of which overlapped heavily with what the nascent self-driving industry needed at that moment.
And so, I chose to go down the route of mapping and localization. Which simply means helping a robot figure out where it is with respect to a 3D or 2D map of the surrounding environment. One may map out the environment in advance or when needed. Usually, it was a combination of the two.
I enjoyed the problem, I enjoyed being on the prow of a fast-moving ship. I was pretentious enough to flout that I was working on real robotics while most of my classmates gravitated towards the fields of Data Science and Ad-tech, lured by the promise of high returns and highly paid immigration lawyers. Turns out that they were the smart ones.
Maybe I would tell the grad-school-me to generalize and not specialize. Actually, that would be solid advice to any current computer science student. With the break neck speed of AI advancement, disregard for safety rails, and abundance of flighty investors, the high-tech AI industry is far from stable. Gen AI is at the bull end of the hype cycle at the moment, but sooner or later even the GPT gods would have to come back down to earth.
But never mind my own story, let’s talk about self-driving vehicles!
A Short history of AVs1
Many of the early self-driving startups sprouted out of the DARPA sponsored Grand Challenges set in the deserts of the Southwest in the mid-2000s. To think that an industry was born out of the race to make a autonomous killing machine should make any AV-hopeful pause. But never mind the icky ethics, let’s move on.
The companies that were born out of this challenge converted their entries into early prototypes and ported over much of the coding stack into their commercial ventures. The resulting corporate cultures were built around the myth and inventiveness of the early founders with an equal amount of excitement and fanfare surrounding the inevitability of a self-driving future.
On my first day at the office, one of the senior execs showed us a sparse set of slides detailing the product’s capabilities as well as its potential to help the people. “Which people?” I never though to ask. I assumed that we had the same people in mind.
The start-up I was working at folded in its fifth year, leaving behind shocked workers to deal with an abrupt aqui-hire to a Tier-12. The Robo-Taxis we worked on were unceremoniously scrapped and sent to the junkyard to be smashed into thick layers of steel and aluminum. Watching the years-long work of countless workers get pummeled was a difficult end to the tale. This company had been one of the slightly more planet-and-people focused ones with EVs and plans of solar grids and driving at slow speeds within geo-fenced areas. Unfortunately, even scoping down the problem was not enough to save us from a slew of poor decisions. Most of the engineers and the management jumped ship at the first chance they got. Never mind that some of their new vessels proceeded to sink soon enough.
Sadly, the waters were choppy everywhere. And our story was not unique.
From mid-2021, many self-driving startups got caught in the same vortex. And those of us working within the industry really shouldn’t have been as surprised as we were. Obviously, the self-driving problem is a very difficult one to solve. Even a single component of its en-suite sensors, the GPS, took over 30 years to get to a commercial state, and it still fails under common road geographies like bridges and tunnels.
The reasons for the industry’s current flailing are easy to grasp: high ideals, terrible forecasting, and a dash of technochauvinism. Add in the hype and rabble-rousing tactics employed to raise capital and you have a pretty good idea of what happened. From the beginning, impossible deadlines and futuristic features dreamed up by executives who operated far from the test tracks put an entire industry on the hunt for a white whale3.
Investors of the 21st century - private equity, traditional auto, the Koch family, and other mysterious entities - were blinded by Zuckerberg’s ‘move fast and break things’ rule. However, building a self-driving vehicle is not the same as building an HTML page deployed to rate women.
The AV industry inched along instead of flying at a break neck speed. And so, investors being investors began closing up their wallets and moving on to the next big thing.
A Brief Glimpse into The Situation Now:
The companies that went on to achieve the holy grail of public listings like Embark, TuSimple and Aurora, saw their share prices tank from their initial offerings. Embark later got acquired by Applied Intuition, one of the bigger players in the AV space. In Jan 2024, TuSimple delisted from NASDAQ and is now quietly selling off its trucks in preparation of exiting the US.
Many unlucky startups have shut down with no reprieve within the past couple of years.
Some other startups that chose the route of buyouts and acquisitions find themselves in a sticky situation yet again. Several OEMs that invested in such offshoots devoted to the self-driving mission have been actively slashing staff and shuffling the rest into existing projects within their vast bureaucracies. Even Motional is facing a change in its winds after Aptiv pulled out last month and is forced to now cut 5% of its non-tech workforce.
And finally, poor Apple should have stuck to improving its products instead of taking on two momentous challenges: building its own EV AND developing self-driving tech. A precautionary tale for proper product development with an unfortunate ending for its remaining 1400+ employees (previously ~5000). According to TechCrunch, a few favorites will get absorbed into the GenAI vertical, while others have 90 days to find suitable roles within Apple by themselves. Which means that, of course, there will be layoffs. Again.
As always it is the workers that suffer. I know many hard-working, smart, and wonderfully kind individuals stuck in the throes of this wayward industry. Many of them are specialized knowledge workers on visas who are now living with the stress of constant uncertainty. An ex-colleague said that her newest hobby is cramming in some LeetCode and spending hours pouring over old textbooks and projects. The Tech Job Interviews have gotten notoriously worse after all.
In these situations, it’s important to take a step back and ask WHY? Why are all these talented people being thrown into the sinkhole? Especially when automakers are posting record profits since 2021. The same time period, yes.
Ford, Google, Volkswagen, Tesla, Apple: They are all doing more than okay. Even GM, despite the auto worker strikes and bungling of billions by Cruise. These companies continue to generate record amounts of revenue, while the money saved by the rounds and rounds of layoffs look impressive only to their shareholders. The NASDAQ has been hitting highs in recent times4, thanks to NVIDIA pulling along some of these companies for the win.
What can We Do?
I don’t know if there is one single solution for the people who remain in the self-driving industry. But it can be helpful to keep an eye open, an ear to the ground and to question the workings of capital. I’m glad I decided to quit the industry, as I cannot imagine working for these vultures ever again. On the other hand, I really hope my friends and ex-colleagues manage to maneuver themselves out of the tight corners they find themselves in. Or maybe its time white collar workers shed their differences and joined the picket line.
Currently, the industry is complex, convoluted and highly fragmented with different companies racing to provide different parts of the solution, building on biased datasets and hard-coded planners. Perhaps, if from the beginning, there had been more knowledge sharing and less involvement by VCs we could have found an equitable solution. More importantly, one that works and doesn’t drag along innocent pedestrians.
As for the intentions and motivations behind many of the start-ups - last mile and accessible mobility solutions - it would have been wiser, perhaps, to develop cheap electric public transportation fleets instead of relying on oil-and-auto money.
AV: Autonomous Vehicle
There are three main levels to the Auto/Self-Driving industry:
Full Solution Providers like Tesla and traditional OEMs (cough Apple) who produce both the car and the tech.
Tier-1 suppliers that provide bits and pieces of the hardware and software needed to make vehicles but do not manufacture cars themselves.
Tier-2 are Sensor and Hardware Suppliers: the ones who provide the ability for a vehicle to see and for software to compute.
I was X-Ray reading Moby Dick recently which led to the slew of references in this article. Apologies!
As of the first week of March 2024.
Thank you for providing this comprehensive image of autonomous vehicles and, above all, for entering various data and providing support. I also really appreciated the distinctions between the different types of autonomous cars! (I subscribed!)