The Argo AI Diaries Part 1: How I Crashed My Ford In The Garage Of A Self-Driving Car Company
A lesson in hubris for the founder of the Human Driving Association is a teachable moment for us all.
"I've got bad news," I said to Mike Levine, Ford's North American Product Communications Manager.
"It's not that bad," I said, "but...let's just say the news is...suboptimal."
"Mike, I crashed your truck."
"No worries. Shit happens. Is anyone hurt?"
"Where did this happen?"
"In the garage. At Argo AI."
THREE WEEKS EARLIER
Everything was going perfectly.
I — the founder of the Human Driving Association, infamous critic of BS around autonomous cars and "New Mobility", blistering foe of those who confuse Tesla Autopilot with self-driving, and merciless enemy of "Innovation Experts", "Radical Disruptors" and anyone with "Speaker" in their LinkedIn profile — was about to announce my new role as Director of Special Operations at Argo AI, a deliberately vague position at one of the world's leading self-driving car companies, and the only one with the courage to bring my Brobdingnagian devil's advocacy in-house.
There was only one problem. I needed a real car to use in Pittsburgh, at least for a few weeks. I just had a baby. I needed to lug stuff around. Argo's headquarters is in Pittsburgh, 373 miles from my home in NYC. That's 5 hours in a car I own, or 6 hours in one I don't. And nothing — not my Cannonball Run-prepped 2000 BMW M5, my unheated 1987 911 Targa, and especially not that rolling-casket-also-known-as-a-Morgan 3-wheeler — was going to survive a Pittsburgh winter. Argo may be building self-driving cars, and yes, Argo received a $1B investment from Ford, their first customer, but Argo doesn't have free extra cars lying around, human or self-driven.
There was only one solution. I needed to borrow a press car.
I like to say I'm not a journalist, but boy, have I enjoyed some perks. Not the kind of perks that require whoring myself out. If you want to get invited on junkets, you have to bend. The best junkets require kneeling. That's where automotive "journalists" and I differ. If you've noticed a shortage of car reviews — let alone sneak peeks — from yours truly, that's why. The if-you-can't-say-anything-nice rule doesn't work for me. Instead I've tried to follow the Chris Harris Rule: develop unique expertise, tell the harsh truth long enough, and eventually manufacturers will come back around. Maybe not all of them, but at least those with sufficient confidence in their products to grant a critic access. My expertise? Long (ahem) distance testing of semi-automated systems like Tesla Autopilot and Cadillac SuperCruise. ADAS — or Advanced Driver Assistance Systems — are poorly understood even by owners, and conveniently installed primarily on the higher end cars I liked to borrow.
Who would lend me a car for two weeks? Two weeks is a long time for a press loaner. If you can get a Tesla, four days is a stretch. Better to save that call for another EV Cannonball record. Porsche? They don't really have ADAS worth testing...yet. BMW? Same issue. Mercedes hasn't shown me any love since I eviscerated Drivepilot and its 2016 marketing campaign. Cadillac? I've already covered SuperCruise in the CT6. What about borrowing a car one isn't going to review? Beneath me. Besides, I needed space, AWD and something that would look cool dirty. Volvo? I'd just borrowed a stunning V90 wagon with the latest Pilot Assist, and had a review in the pipeline.
Ford? Too obvious. Once my Argo AI job was made public on January 14th, everyone would assume that the Ford/Argo AI relationship meant a positive review was bought and paid for, and a negative review was done for effect. Besides, I'd already borrowed a Raptor — indisputably the world's best truck — and my review was weeks overdue. I'd never get that out before the 14th, and now even that review would be suspect.
And Ford did not yet know that Argo AI CEO Bryan Salesky had hired me. Even if they did, I couldn't ask for a vehicle on that basis alone. That would be presumptuous, rude, and exactly the kind of thing I eviscerate others for.
So I called my old friend Alan Hall, Ford's Senior Communications Manager for Autonomous Vehicles. He was always down to reject a bad idea.
"I've got a bad idea," I said. "You should lend me something with CoPilot 360."
"Is this because you need a car for something suspicious, or for a real review?"
"A real ADAS review," I said. "I hear it's pretty good, as far as ADAS goes."
"Who told you that?"
"You're about to tell me. And then you should tell me you're so confident in it that you want to lend me a truck with it, because I just had a baby, and I'd love to do some endurance testing."
"Yes," said Alan, "CoPilot 360's good. We just don't overpromise like—"
"You know who?"
"You KNOW who," said Hall. "But you're not driving cross country again, right? With a baby this time?"
"Of course not."
"Can I get that in writing?"
"Yes. But I AM thinking of taking a road trip to Pittsburgh."
"And why's that?"
"Visit Carnegie Mellon," I said. "Maybe do a story."
"You should visit Argo AI while you're there."
"Maybe I will."
"When's the last time you damaged a press car?"
"Never. Come on, man. I'm the founder of the Human Driving Association. It's my job to set an example."
"Call Mike Levine," said Alan. "Maybe he has something."
Levine is exactly as important as the title of North American Product Communications Manager for Ford Trucks suggests. The Ford F-150? Most popular truck in America. Some say other trucks are made for babies. The Explorer? In his portfolio. The Raptor? Duh. Levine was a great friend to have, but I was still working on the friend part.
"I've got an Expedition," said Levine, "or an Edge."
I'm not much of a truck guy. As a driving enthusiast, trucks scare me. Not other trucks hitting me, but me, driving a large vehicle with a high center of gravity and limited sight lines, hitting everyone else. I've spent my life driving small cars as fast as possible. I trust my training, experience and instincts to keep me safe in a small vehicle. A large one? Not so much. I've been driving my wood-framed Morgan around NYC for six years. And I was still alive.
"The Edge," I said, "that's the crossover, right?"
"Yup. I've got a brand new one with the Titanium package, CoPilot 360, fully loaded."
And everything was perfect. A lovely brown Ford Edge was delivered to my apartment in NYC on Monday, January 14th, the same morning my Argo AI position was announced. Best wishes flowed in from unexpected corners, including rival self-driving companies, and even some top people at Ford.
Now...I'm not suggesting Ford CEO Jim Hackett was referring to me in his presentation at the Detroit Auto Show. But what I am saying is that others think he was. My stock was rising. Then Amnon Shashua, CEO of MobilEye, one of the world's largest vehicular camera companies, followed me on Twitter.
Was I getting a big head? Yes. It was time to pack my MFing free Ford Edge full of fake German Polizei jackets, scarves, Quest bars and get on the road. The 373 miles to Pittsburgh? No problem in the Edge. It had all the features I was looking for: room for seven empty coffee cups, nooks for prescription Ray-Ban aviators in pink, blue, yellow and brown, and a cellphone cradle superior to that of a Mercedes S-Class. I didn't know an affordable crossover could be so comfortable. Coming out of an eighteen year old M5, the Edge was downright luxurious. I'd have to wait for the impending storms to see what it could do. CoPilot 360? I'd break out the GoPros on my first weekend in Pittsburgh. I was within 24 hours of a new career as an executive at Argo AI, answerable directly to CEO Bryan Salesky, poised to learn secrets about which I'd been guessing for four years, working at the very heart of the automotive future.
What could go wrong?
It would only take three days to find out.
In their generosity, Argo offered me a prime parking spot in the basement below their Pittsburgh office. Although New Yorkers are among the world's best parallel parkers, there's a vast difference between parallel parking a Morgan 3-wheeler — tiny and low to the ground, with 360 degree visibility — and parking a crossover — with limited sight lines further blocked by my delusions of grandeur — just right of a big pillar.
Every day for three days I'd slowly navigate the garage counterclockwise, miss the optimal turn-in point, and lose an entire thirty seconds backing up and re-entering my spot at the correct angle.
I was Argo AI's Director of Special Operations. I didn't have time for this.
On my fourth day of work, I had a plan. I was going to arrive before anyone else, navigate the garage clockwise, make a wide right...then turn left into my spot, saving an entire thirty seconds.
It was a perfect plan—
Let's do a little accident reconstruction.
What's that? A low pillar on the right, just below the Edge's belt line?
But wait! What's this?
Yes, I had completely not noticed that I was going the wrong way against traffic.
The situation was suboptimal. I hadn't so much as scratched a car in 15 years. I'm supposed to be the guy who nails the crazy stuff. I'm supposed to be the guy who makes the unsafe SAFE, not the safe dangerous. Alex Roy. A Ford Edge. Crashed. At 5mph. In the basement. Of a self-driving car company.
I was ruined.
Even in shame, honesty is the only policy. I called Mike Levine.
"Mike, I crashed your truck."
I told him the whole story. Mike was as understanding as the North American Product Communications Manager for Ford Trucks could be, which is to say it didn't sound like I'd be borrowing any more Edges, let alone Raptors. I had only one option to salvage the relationship.
"It's my responsibility," I said. "Send me the paperwork."
"Have it your way."
And then it dawned on me. I wasn't ruined. My professional reputation wasn't weakened. It was strengthened. This was a gift. From destruction comes creation. My skills were atrophying right before my eyes, and with them the fantasy that I could somehow stave off some form of automation with nuance, clever writing and more Skip Barber classes. At least some form of automation isn't merely inevitable, but necessary.
Let's get real about autonomy. Not autonomous cars, but autonomy itself.
If autonomy = freedom, what is the Ultimate Autonomy Car?
It's not a geofenced self-driving bus or taxi, which are coming whether we like it or not. It's a Ford Raptor (or equivalent) with:
- A self-driving option for city/highway driving that's demonstrably safer than the average human.
- A parallel automation system — think ADAS 2.0 — that mitigates/prevents on-road crashes while under human control.
- An off-road mode with zero automation and 100% human contol.
- A privacy button to shut off all connectivity.
- A self-parking feature for idiots like me.
Thats's MY perfect feature set.
I'm no Ayrton Senna, and never claimed to be. But I always thought I was "safe", if not way better than the average driver. That may be true, but I still hit that pillar. In that moment, driving a popular crossover, in conditions similar to what millions around the world experience every day, I failed. I was lucky there wasn't child instead of that pillar. There's no excuse for it. But there are two good answers. We need better systems to help humans control machines where machines are incapable of taking full control. That's called parallel automation, and it's what you find in an Airbus, and where Toyota is trying to go with their Guardian systems. We should also embrace safety when and where we can, as long as it augments human freedom rather than limits it. That's L4 automation, and it can't be deployed until we — all of us — agree on what "safe" means. The self-driving sector has a lot of work to do, but so do humans.
For example, why should Mike loan me a Raptor in the future?
Because I'm willing to take responsibility for my mistakes. I hope Mike thinks that's good enough.
(As for that Ford Edge review: Love it. Terrific. Amazing. Could use more plastic body cladding, especially at waist level.)
(CORRECTED to reflect Mike Levine's initial comment upon learning of the crash.)
Alex Roy is Director of Special Operations at Argo.AI, founder of Geotegic Consulting and the Human Driving Association; editor-at-large at The Drive; host of The Autonocast; co-host of /DRIVE on NBC Sports; and author of The Driver. He has set numerous endurance driving records, including the infamous Cannonball Run record. You can follow him on Facebook, Twitter and Instagram.