This is the first time I saw anything about this. It’s different than Caltrans + UC Berkeley we’re doing as they used RTOS and road sensors, but this seemingly is mostly using cameras to assess navigation. It’s pretty neat for the time albeit in quite predictable traffic.

He says he started in 1977 on the vision system.

https://ethw.org/Oral-History:Ernst_Dickmanns

“ And of course, what you could observe was at every PhD generation, I say 4-5 years, the computing power increased by a factor of 10. So within 15 years, we started 1977, from '80-95, this is a factor of 1000 in computing power. ”

We’re at 1 million times in computing power by now?

> We’re at 1 million times in computing power by now?

And according to a friend of mine with expertise in GPU programming, Tesla's HW3 chips still simply aren't powerful enough to run a sufficient NN for real-time world scene reconstruction from streaming video (indeed, the demos that Tesla has done were using water-cooled supercomputers sitting in the trunk/cargo-area). But here's to hoping!

I wonder whether in 1977 the AI community told him not to bother because AI is going to solve it anyway. http://www.incompleteideas.net/IncIdeas/BitterLesson.html

This is just a restatement of Moore's Law (doubling every 18 months). 2^(5/1.5) = 10.1.

I wonder how much of the research and technology from that time has trickled into today's tech?

When I was in college in the early 2000s, the DARPA challenge was all the rage. I feel a lot of the r&d from that time had direct influence on today's tech.

Semi autonomous vehicles have been around for a very long time... But the tech still has a ways to go to get to full autonomy.

It's surprising how similar it is to modern Teslas, forty or so years later.