State Of Autonomy: May Recap

Screenshot of Jalopnik.com’s top stories, a day in mid-May

Every month, I recap the news articles I’ve consumed around autonomous vehicles, calling out the highlights and keeping track of market projections. This is also your chance, dear readers, to nominate a topic for discussion in the following month.

The preliminary NTSB report on Uber’s fatal Tempe crash came out this month, and while it provides the first actual information that matters to anyone who cares, it probably received less than 5% of the exposure that the crash event itself did. Typical media cycle, typical talking heads… typical irresponsible behavior all around. Consider yourself lucky if this is the first you’re hearing of the fatal Uber collision. And, if it is, I recommend these two articles:

• EETimes: a good breakdown of the timeline of events before the collision, where the failures occured, and a gem about safety cases from Phil Koopman thrown in for good measure
• Brad Templeton: a more detailed assessment of the failure points, and some well-informed speculation as to the hows and whys

If you’d rather get the 30-second digest:

• Uber’s perception software (think of this like the relationship between your five senses and your brain) wasn’t mature enough to accurately classify what it was sensing; given the circumstances and possibly some programming decisions we are as yet unaware of, it did not prioritize the victim as an obstacle to avoid until 1.3 seconds before impact.
• At -1.3 seconds, the software likely would have applied emergency braking to avoid the victim, but Uber had previously disabled emergency braking due to its past performance delivering a lot of false positives (again, a matter of software immaturity), leaving the task at this point up to the safety driver.
• At any point in time, the safety driver could have performed an evasive maneuver to avoid the victim. Notably, there was only one Uber employee in the vehicle, whereas Uber had until recently used two operators to share the duties of road monitoring and vehicle monitoring. Two operators are standard in every other company’s testing environment, as far as we know.
• All of the failures point back to Uber management, which likely explains why the company shuttered its Arizona AV testing operations; whether that ejected the guilty parties or merely created a convenient scapegoat to shield upper management, we don’t know. Uber still runs public tests elsewhere.

In unrelated but useful news, SoftBank is a fucking monster. The Japanese firm who was once known for telco has since become infamous for having put together a $100 billion dollar tech fund (The Vision Fund), and cutting life-changing checks to startups with said fund. The latest was GM’s Cruise, who announced a $2.25B investment from SoftBank and a long-term commitment as partners. Since GM seems hell-bent on developing an actual self-driving car, this investment would imply that SoftBank is making a (let’s face it, small) bet on Cruise to deliver the goods in a timely manner. That would indeed be great for SoftBank, who also owns a stake in pretty much every other component of the AV stack — from rideshare giants to trucking to chipmakers to mapping software to pureplay robotics. This firm is the quietest and most unique Google rival on the planet.

This Month’s Highlights:

• SoftBank Injects GM’s Cruise With $2.25B Investment; GM Allegedly To Invest An Additional $1B
• Waymo Plans To Purchase “Up To” 62,000 Chrysler Pacificas For Robotaxi Commercialization Beginning Late 2018
• NTSB Report Shows Uber’s Fatal Tempe Crash Was A Perfect Storm Of Immature Software And Disabled Redundancies
• Uber Shuts Down Its 300-Person Arizona Self-Driving Test Program
• Waymo Expanding Its (Fully Driverless) Arizona Robotaxi Service To Phoenix Later This Year
• Aptiv To Bring 30-Car Robotaxi Fleet To Las Vegas For Operation On Lyft’s Rideshare Network
• Drive.ai (AI Research Legend Andrew Ng’s Company) To Launch Small Robotaxi Service In Dallas Suburb, Beginning This July
• Ohio Governor Signs Pre-Emptive Executive Order To Support Autonomous Vehicle Testing In-State
• Chinese Self-Driving Startup Roadstar.ai Raises $128MM Investment One Year Into Its Existence As A Company
• Smart Pavement Startup Integrated Roadways Lands Contract With Colorado DOT

Market Predictions:

Note: think of this as the ugly transition between the chart I’d been previously offering and a more usable but not-yet-fully-baked visual to come. I really aim to inform people with this thing, but it’s difficult to achieve without making some major assumptions.

Also note that I don’t list every Tom, Dick and Harry who makes a prediction about their own progress — if you want to see what that looks like, check any of my monthly updates before 2018. For instance, you don’t see Uber here (for reasons that should now seem obvious), nor do you see Fiat-Chrysler who just made a baseless claim to their investors that they’d have Level 4 autonomy in 2023. I hand-picked a group of the more serious and significant players from multiple disciplines and geographies. I will continue the chart in this manner, although again, it requires assumptions. Trust me: you don’t want to see a chart that contains no assumptions.

Coming In June:

• Instead Of Ten-Figure Funding, This Robotaxi Startup Has Actual Customers
• Your suggestion? Send a tweet to Mitch Turck

[Update 6/18]: You’ll now see me writing contributor articles for Forbes, many of which will also be posted to Medium. The good/bad news is that I’ll be ranting briefly several times a month instead of hemorrhaging one or two 5,000-word essays. You can follow the Forbes antics here.

Reactions From The Public:

Re: Everything That Scares Us About Autonomous Cars Is In Windshield Wipers (My Article)