Of course, you and I are pretty much on the same page. The complexity and cost would be staggering to redesign the whole thing. But maybe there are smarter, incremental ways to go about it. Perhaps mandate a special paint for lane markers to include magnetic paint or some kind of nanotech. We already do repairs on roads, repaint, replace "cat's eyes markers" etc, so modifying the materials used should be easier.
I fully agree it's going to take a long while for everyone to adopt this, and one of the keys will be proving the AI tech - which then becomes a chicken and egg problem. Maybe Google/Tesla are laying the groundwork for that, and on second thought maybe I'm wrong about us thinking about this the backwards. Perhaps this "over-engineered" solution (for the lack of a beter word) is exactly what we need to then later optimize for "the right way".
I think about the half-measures like new paint for lane markers and I'm equally skeptical, but that's probably just me being pessimistic.
All we really need is an idea that we can't fathom today. Google has a library of images of just about every street in the states. If they could run that same project and in the process lay out markers of some sort that would be helpful for the next generation of autonomous vehicles, then from that point forward incremental change is simply a matter of re-tooling an autonomous vehicle. Cooperation from local governments could be pushed through in a similar way that the 55 MPH speed limit happened.
Of all the "billion dollar ideas" floating around silicon valley, the person/investment group who's willing to take on an idea like this to a nationwide scale is going to be the winner.
The first round doesn't even need to be high tech. They could have a paint ball gun hanging out the side of a car spraying out a biodegradable path that a series of other cars can follow. Those cars could then train other cars.
...
Yeah, I'm flip flopping. I'm actually an idealist deep down inside. Sometimes I forget that for a little while.
I fully agree it's going to take a long while for everyone to adopt this, and one of the keys will be proving the AI tech - which then becomes a chicken and egg problem. Maybe Google/Tesla are laying the groundwork for that, and on second thought maybe I'm wrong about us thinking about this the backwards. Perhaps this "over-engineered" solution (for the lack of a beter word) is exactly what we need to then later optimize for "the right way".