Tagged “cars”

Self Driving Car Update

About 10 years ago, I got up on a stage to give a lightning talk about self-driving cars. If you want to relive that person I was, it’s still up on Youtube (and my talks page, so I can’t be accused of hiding stuff). Here’s an embed:

Now, the idealistic person I was in that talk made a fair number of points that I still generally believe:

• Dedicating a ton of space to cars that are not in active use rather than human-centric infrastructure is bad.
• We should, as a society, work to reduce traffic deaths - especially those involving a car and a not-car (pedestrian, bicyclist, and the like).
• Driving is a chore, and at the time I sorta resented living in a place where it was an abject requirement, both for me and folks who really would have been better served by on-demand transit.

At the time, I was pretty bought into the hype that we were making great strides in self-driving technology. I even at some point after that talk repeated the talking point of “we’re just five years away” (because we’re always just five years away). A lot of that idealism was born of a few different things: Tech was still a generally fascinating and exciting area of endeavor for me and I was still very optimistic that the industry generally wanted to make people’s lives better. Likewise, I lived in a place with horrendous public transit. Very few bus lines, 1 hour wait times, just generally unusable. Cars are a necessity there - and cars that can park themselves or be summoned on demand seemed like the best option.

I do want to give Oklahoma City a bit of credit here, their transit system has gotten a lot better in those 10 years - still not where I’d expect a major city to be, but it has markedly improved.

Likewise, there were a lot of demos being shown at the time faking cars abilities to drive themselves (Tesla, for example, in 2016 a couple of years after my talk - at the time it was the Google Incubator precursor to Waymo and Cruise doing demos).

I don’t really think these things anymore - and part of that is a change of scenery. I now live in a place and have visited places that have great public transit regularly. While trips can still take a lot longer than a car would, it’s not nearly as big of a burden as it used to be.

Sidebar: We do still have plenty of work to do, as always.

One of the things that I dismissed as impossible (because of my environment) was that we could build effective transit across the country and in places where perhaps the political will wasn’t very present has been replaced with a sense of displacement. The displacement is the persistent need from these automation companies to convince cities not to invest in other modes of transportation because why would you if the automation revolution is right around the corner. All that investment would be "wasted" (it wouldn't), so it behooves the automation companies to continue to say things are "just a few years away" for at least a decade, on repeat.

And like, progress is being made. Waymo is running a relatively-sanctioned experiment in a couple of amenable cities. Tesla is running an unsanctioned experiment with every driver who pays them more to do so (which has resulted in plenty of fatal crashes). But hey, these cars are running all on their own... or are they just offshoring the driving to remote workers? Hard to say for sure when the AI companies are constantly faking their metrics and demos. Long story short, they have lost the benefit of the doubt. They perhaps should never have had it in the first place, but it was a simpler time. Or at least it was (to me) without another decade of being in the tech industry.

Instead, I have personally doubled down on building out robust regional transit and investing in high speed cross-country and regional rail to cover our moving-people-from-one-place-to-another needs. For freight, I think we should focus on making working conditions for truckers the best they can be and maybe accept slightly longer supply chains (I think that we're going to have to deal with that regardless, we've already experienced several "historic" disruptions all in a row and climate change is only going to make it worse).

We need more bike friendly cities and more well connected suburbs. We should invest in transit at scale, it's a proven and time-tested thing.

Could we do that and create autonomous vehicles at the same time? Sure, yeah. However it's going to be very difficult to do in a world where politically we prioritize one thing over the other and defer to the private sector on too many things. If we were to do that, I'd also want something akin to a smart grid where the cars could know where each other are at all times, because otherwise you get scenes like this:

Endless cars, endlessly unaware that the other cars are also on the same system.

But such a system would also take us even further down to a dystopian surveillance state where your non-smart car is also tracked or you're encouraged to install a system to cover the tracking. Yeahhhh, maybe not great.

Anyway, in a world where we can't have both, I'd much rather have the good transit than the "maybe it works maybe it doesn't" promise of "just give us another 5 years and we totally go it, trust us".

So here's another thing that's a little bonkers about self-driving cars, invariably someone will say "Try it! It works great." And yeah, it just might! For you. For that one drive. What about when it holds up traffic? What if it breaks a traffic law and there's no one to ticket?

What matters is "does it work at scale" and "did we have to kill anyone in the process"? Waymo is probably being the most socially responsible one here, it's getting permission before deploying its fleet in pilot cities. Meanwhile Tesla's out there running an experiment on every road everywhere but prioritizing its CEO's drive.

Now, I imagine you saying "it just needs to be better than humans"! Sure! But by what metric? Humans know how to navigate construction without seeing that in their training data a whole bunch of times. And by who's account? If the data's all being held by the company and not being audited independently... who can be sure it's actually safer?

All of these metrics are currently vibes based. "It will get better! It just needs more data" is a vibe. "My Waymo ride was awesome" is a vibe. We're rolling out a consequential technology based on vibes and just letting huge US companies cop out with a "trust me bro".

But, the NHTSA is investigating currently as I write this, so we'll see what they find.

No. I hate being lied to about progress. I hate technology that isn't centered in our needs. I hate tech that's just being used to enrich a few giant companies.

That's why I don't think we're going to see large-scale self-driving any time soon. And that's okay.

19 Aug 2024 by cthos

blog AI cars civics