Honesty is the best policy, but it’s not company policy.
Years ago I wrote about the cheating of NASCAR car drivers. And recently at the last BSidesLV conference I pointed out in my talk how human athletes in America get banned for cheating, while human car drivers get respect.
Anyway I was reading far too much on this topic when I starting thinking how NASCAR studies of ten years ago to end cheating could be a compelling area of research for ethics in driverless cars:
Proposed solutions include changing the culture within the NASCAR community, as well as developing ethical role models, both of which require major action by NASCAR’s top managers to signal the importance of ethical behavior. Other key stakeholders such as sponsors and fans must create incentives and rewards for ethical behavior, and consider reducing or ending support for drivers and teams that engage in unethical conduct.
That’s some high-minded analysis given the inaugural race at Talladega (Alabama International Motor Speedway) had a 1969 Ford with its engine set back nearly a foot from stock (heavier weight distribution to the rear — violating the rules).
This relocation of the engine was easily seen by any casual observer yet the car was allowed to race and finished 9th. Bill France owned the car. Yes, that Bill France. The same guy who owned the track and NASCAR itself…entered an illegal car.
An illegal car actually is icing on the cake, though. Bill France built this new track with unsafe parameters and when drivers tried to boycott the conditions, he solicited drivers to break the safety boycott and issued free tickets to create an audience.
“I really admired that he told everybody to kiss his ass, that that race was going to run,” Foyt said.
The sentiment of getting everyone together to agree to an ethical framework sounds great, until you realize NASCAR stands for the exact opposite. It seems to have a history where cheating without getting punished is their very definition of winning.
Tech philosophers in America watch closely the attempts of highly-individualistic short-term investor-run truck companies to behave as much on shared infrastructure like trains as possible without realizing the societal benefits of a train
Nvidia boasts in the Sacramento Bee of a truck that was able to drive a load of butter across America
…the first coast-to-coast commercial freight trip made by a self-driving truck, according to the company’s press release. Plus.ai announced on Tuesday that its truck traveled from Tulare, California, to Quakertown carrying over 40,000 pounds of Land O’Lakes butter.
Mercury News says it took three days on two interstate routes (e.g. human life prohibited) and didn’t experience any problems.
The truck, which traveled on interstates 15 and 70 right before Thanksgiving, had to take scheduled breaks but drove mostly autonomously. There were zero “disengagements,” or times the self-driving system had to be suspended because of a problem, Kerrigan said.
Indeed. The truck appears to have operated about as much like a train as one could get, although at much higher costs. If we had only invested a similar amount of money into startups to achieve a 250 mph service on upgraded tracks across America…
2800 miles at 250 mph is just 11 hours. Using electric line-of-sight delivery drones to load and unload the “last mile” from high-speed train stations at either end would even further expedite delivery time.
Trains = 12 hours or less and clean
Trucks = 24 hours or more + distributed environmental pollutants (fuel exhaust, tire wear, brake wear, wiper fluids…)
Trucks can’t improve their time much more because they start becoming more and more a threat to others trying to operate safely on the Interstate’s self-run collision avoidance systems. And it’s exactly the delta between operating speeds of vehicles on the same lines that generates the highest risks of disaster.
The absolute best whey going forward to skim time (yes, I said it) should be clear (it’s trains), although we’re talking long-term thinking here, which investors lurking around startups for 2-year 20% returns on their money have never been known to embrace.
Driverless trucks in this context are a form of future steampunk, like someone boasting today their coal-fired dirigible has upgraded to an auto-scooper so they no longer need to abduct children into forced labor.
Congratulations on being less of a selfish investor threat to others, I guess? Now maybe try adopting a socially conscious model instead.
A talk I was watching recently suggested researchers finally in 2019 had cracked how robots could efficiently act like a swarm. Their solution? Movement based entirely on a light sensor.
That sounded familiar to me so I went back to one of my old presentations on IoT/AI security and found a slide showing the same discovery claim from 1953. Way back then people used fancier terms than just swarm.
W. Grey Walter built jelly-fish-like robots that were reactive to their surroundings: light sensor, touch sensor, propulsion motor, steering motor, and a two vacuum tube analog computer. He called their exploration behavior Machina Speculatrix and the individual robots were named Elmer or Elsie (ELectro MEchanical Robots, Light Sensitive)
The rules for swarm robots back then were as simple as they will be today, as one should expect from swarms:
If light moderate (safe)
Then move toward
If light bright (unsafe)
Then move away
If battery low (hungry)
Then return for charge
Cool kids run their rods on decentralized oil, thanks to Diesel who not only warned of centralization dangers but invented standards-based solutions that work to this day
A month ago I was on a call with some top security experts in the industry. We were discussing my upcoming presentation about exciting control options and data privacy from applying decentralization standards to the automotive industry.
To put it briefly I was explaining how web decentralization standards can fix growing issues of data ownership and consent in automotive technology, a fascinating problem to solve which I have spoken about at many, many conferences over the past seven years.
Here’s one of my slides from 2014, which hopefully increased awareness about automotive data ownership and consent risks:
Much to my surprise I see this issue just hit the big papers for some well-deserved attention, albeit I also see it may be for the wrong reasons.
…data isn’t the new oil, in almost any metaphorical sense, and it’s supremely unhelpful to perpetuate the analogy…
That’s just to frame the many problems with this article. Here’s another big one. The author wrote:
We’re at a turning point for driving surveillance — and it’s time for car makers to come clean…
Haha, turning point. I get it. That pun should have led to “it’s time for car makers to choose a direction”. Missed opportunity.
But seriously, the turning point for many of the issues in this article surely was years ago. He raises confidentiality and portability issues, for example. Why is now the turning point for these instead of 2014 when encryption options exploded? Or howabout 2012 when a neural net run on GPUs crushed the ImageNet competition? I see no explanation for why things are present concerns rather than past/overdue ones.
I’d say the problem is so old we’re already at the solutions phase, long past the identification and criticism.
I had help doing a car privacy autopsy from Jim Mason, a forensic engineer. That involved cracking open the dashboard to access just one of the car’s many computers. Don’t try this at home — we had to take the computer into the shop to get repaired.
Sigh. Please do try this at home.
Right to repair is a very real facet of this topic. Cracking a dashboard for access is also very normal behavior and more people should be doing it.
When I volunteered my own garage space in the Bay Area, for example, I saw the reverse effect. Staff of several automotive companies came to join random people of the city in some good old community cracking of dashboards.
A guy from [redacted automotive company] said “…what do you mean you don’t bring rental cars to take apart and hack for a day? You should target ours and tell us about it.” Yikes. That’s not ethical.
The 1970s “hot-rod” culture in today’s terms is a bunch of us sitting around with disassembled junkyard parts in a controlled garage (not operational rental/borrowed cars on the street!) and our clamps on wires etc to linux laptops deciphering CANbus codes.
This journalist desperately needs to participate sometime in a local car hacking community or at least read “Zen and the Art of Motorcycle Maintenance”….
It should not be hard for a machine owner to crack it open, when market regulations are working right. At least the journalist did not say an “idiot light” forced him to take his computer to the manufacturer for help.
Anyway, back to the point, the data models in automotive need to adopt decentralization standards if they want to solve for data ownership issues raised in this story.
But for the thousands you spend to buy a car, the data it produces doesn’t belong to you. My Chevy’s dashboard didn’t say what the car was recording. It wasn’t in the owner’s manual. There was no way to download it.
To glimpse my car data, I had to hack my way in.
In summary, data is not the new oil, right to repair means healthy markets trend towards hardware access made easy, and concerns about confidentiality and portability of data in cars are being addressed with emerging decentralization standards.
Sorry this article may not come with a viral click-bait title, but I’m happy anytime to explain in much more detail how technical solutions are emerging already to solve data ownership concerns for cars and give examples with working code.
a blog about the poetry of information security, since 1995
