[ricardobayes]

It's a well-known thing in AV circles that some companies were actually faking their route by hardcoding them for demos etc. Also one shortcut is to use HD map data, and relying on 100k+ LIDAR sensor data that will never will be cost-effective. I know of only a handful of companies that rely on 'cheap' sensor data (cameras/radar) and use real-time calculations.

[michaelt]

> relying on 100k+ LIDAR sensor data that will never will be cost-effective

Eh, the $100k prices are mostly for high-end, low-production-volume research LIDARs.

There are LIDAR manufacturers claiming they can achieve prices of $500 [1] or even $200 [2] if they were fitted to every car made.

[1] https://guidehouseinsights.com/news-and-views/lower-cost-lid... [2] https://www.opsys-tech.com/post/are-we-nearing-a-shakeout-in...

[xyzzy123]

If a full time driver earns 25k-50k that's sort of equivalent to every regular taxi and uber having a $500k / $1M driving package onboard already (at 4%), right?

I don't know the industry but very naively it doesn't seem to me like $100k really breaks the unit economics? Particularly if volume / learning curve can reduce the price?

[ricardobayes]

You are right in the way that the industry seems to be shifting from general-purpose self-driving to transportation/hauling.

[papito]

Well, THAT has a Dropout-like HULU miniseries written all over it.