Self driving cars are still decades away

That’s promising.

I was in San Francisco a few days ago and took two Waymos. It’s a very slick technology and in my limited experience as good or better than Tesla on the city streets.

No, Waymo and Toyota did not announce a new partnership.

They announced the intention to explore the idea of maybe forming a partnership.

Is that better than nothing? You bet. Is that a sign of two fast-moving innovative organizations about to roll up their sleeves and get to work making a new future? Hardly.

Toyota has done a lot of these sorts of things over the years:
https://www.wsj.com/articles/toyota-investing-500-million-in-uber-in-driverless-car-pact-1535393774

Granted, one difference here is that Waymo actually has a shipping system. So there’s possibly a bit more meat here.

Yeah. All the big automakers (heck and airplane makers) do exploratory agreements with scrappy start-ups to sniff each other’s rumps looking to see if there’s an advantage to be had. Then if everything smells interesting enough to both they do a deal to actually invest, exchange or license tech, etc. I don’t mean to poo-poo this particular partnering; just the breathless article.

You’re surely right that Waymo is a real product. They may well have cracked the secret to safe autonomous taxis in well-bounded urban / suburban areas. At which point hooking up w Toyota is what gives them scale to deploy everywhere at once in just a year or two while musk is still bragging about what he’ll release next week.

As Bullwinkle so often said: “This time for sure!” Not.

Unrelated to the Toyota thing, I had a question about Waymo. As I’ve mentioned before, I see them pretty much every time I’m out driving around. A few weeks ago, I saw one in the dedicated lane for the freeway. I thought I’d heard here that Waymo’s weren’t allowed to go on the freeway? Because I had to make a turn just as the Waymo was getting to the on-ramp, but it was a dedicated lane it was in and it only could have gotten on the freeway from there. Can they go on the freeways then? Or did I see a rogue Waymo?

I meant to mention that I saw one on the freeway well south of San Francisco. When I got closer I saw a person in the driver’s seat.

Maybe there was a physical driver. I was far enough away I didn’t see.

Rumor has it that someone killed the announcement last month that they exchanged contact information in order to see if they were interested in having an intention to explore the idea of maybe forming a partnership.

As we said in USAF…
Time to prepare to plan to possibly proceed to prosecute with prejudice.

Completely self driving trucks, with no supervising driver, will begin making “driverless customer deliveries between Dallas and Houston.” They claim only 1200 miles of non-supervised driving, but three million miles of supervised autonomous driving.

I-45 next week:

Or…

Uber back in the game. Arlington, Texas, “later this year.”

https://investor.uber.com/news-events/news/press-release-details/2025/Uber-and-May-Mobility-Announce-Strategic-Partnership-to-Scale-Autonomous-Vehicles/default.aspx

Reading the article, they use AV for autonomous vehicle such as AV market.

Here in Japan, AV comes from “adult video”
which what they use for porn.

The rather dry news release is more interesting is you substitute “porn” for AV.

To me AV is more commonly used as an abbreviation for audiovisual or anti-virus.

I was in the AV (audiovisual) industry before.

Hypothetically, one could make true but intentionally misleading statements about being in “the AV industry”
when meeting people at parties or bars but I’m sure I was far too mature to stoop to such antics.

AV is porn for capital-rich executives. Works for me.

Hmm, i have an “AV” folder on my computer. It’s where i keep my photos, mps files, videos, and other, um…

Should i be worried about getting raided? :laughing:

To counteract the title

I was thinking this morning that a single self-driving scale doesn’t really capture all the detail you want about a system’s performance. Two axes (like a two-axis political scale) might work a little better.

Suppose you divide performance into Roads and Driver. Roads scores like:
1: Only supports limited roads in a limited area. (GM Super Cruise, Aurora trucks)
2: All/most public roads within a limited area. (Waymo)
3: All/most public roads within a wide area. (Tesla FSD)
4: All roads in good condition, public or private.
5: All driving surfaces of any kind that a human can drive on.

And then Driving:
1: Only limited driving assist functions available; driver is still in the loop. (Most lane assist tech)
2: Car supports most functions but driver must be ready to intervene at any moment. (Tesla FSD)
3: Car can handle all functions or degrade safely. A driver does not have to look at the road but must be available for some situations.
4: A driver doesn’t have to be present. There may be a remote operator available for some situations. (Waymo)
5: No driver or remote operator is required. The car is fully autonomous.

Tesla FSD is R3D2, maybe bordering on R4D2 (it can handle some driveways, parking lots, etc.). Waymo is R2D4. Geofenced to a few cities, mostly surface streets only, but no driver. Does need a remote operator. Best I can tell, the Aurora trucks are R1D4, since they’re localized to just a couple of highways and access roads.

Not sure if there are any R2D2 robot drivers at the moment.

A 5 on either scale basically requires AGI. For roads, it implies that the car can handle arbitrary 4WD scenarios. And for the driver, it implies (for example) a language model, so that one can command it to handle arbitrarily complex situations, as well as behaving itself in unanticipated situations, like doing something safe after an accident.

I’m sure it’ll be a while, decades maybe, before we have a R5D5. But an R4D4 is probably coming fairly soon.

Ehh, I think you’d need another axis for weather conditions. I’m not sure if that would be one axis or two, though. Snow, fog and rain all require different capabilities from the driver.