>
I specifically said that it should be done on high volume, maintained roads, such as highways and ring roads. It doesn't have to happen overnight, but it could happen gradually. It would serve other purposes too, like guidance for snow plows. It's a 80/20 solution - cover 80% of the travel for 20% of the cost. Solving the problem on highways and high volume roads without cyclists are significantly easier, and would still give a tremendous benefit.I've travelled by road a fair bit around different countries and I assure you that a lot of high volume maintained roads still fall into the categories I outlined. Moreover, a lot of high volume roads aren't maintained. I honestly don't think you would even get remotely close to covering 80% of the roads for 20% of the cost. Maybe if you isolate specific districts like Silicon Valley then your figure might be accurate. But for "normal" towns in Europe that wouldn't be the case and, as has been discussed elsewhere on HN, having some roads updated and others not would actually be more dangerous than having none redeveloped because humans would demonstrate the same complacency across all the roads once they're used to autonomy on the new roads.
Also highways aren't the problem - they're significantly easier to determine in code and see fewer deaths per year. It's the smaller streets in towns that are the problem. Both from the perspective of statistical deaths and also from the perspective of the complexity of AI required to navigate them. They're what really need guide wires the most.
However going back to highways: even on motorways you still have the problem of motorcyclists weaving in and out of traffic. So I'd argue that you haven't entirely eliminated the "cyclists" problem - you just now have faster and even deadlier ones.
I should say that I do agree with you that there would be tremendous benefit if it were feasible. What I disagree with is just how feasible it is.
> Regarding laying the cables in existing roads. We're already pretty good at "shooting" cables under existing roads with pneumatic solutions, and I'm confident that we could do so for rural roads at a modest price. But that's a different talk.
Shooting the cables is the easy if you have pavements with cables already installed. If you don't then you need to dig up the road to lay them. And that's still only half the story. There's a whole load of costs involved in the planning stage; scheduling the work around busy periods (which sometimes means paying for night shift work) and organising the road closures. And it's the maintenance of that infrastructure after the work has happened. Remember that a lot of busy streets aren't even maintained currently. And it's doing that work sympathetically around historical locations (where applicable).
There's a significant amount of more cost and work involved than I feel you're giving credit for.
> If we want to develop an "AI" for this, it's not going to be ready in 10 year, nor 20 - we do not have the processing power, or understanding of the domain to do it at this point, and we're moving rather slowly.
I completely agree that we're decades from a solution but I disagree that we're moving rather slowly. 10 years ago the stuff that are in consumer vehicles - even some entry level cars - would have been considered Sci-Fi just 10 years ago. But even if we are 50 years away, I can't see cables laid in twice that time. I mean half the towns in England and America don't even have broadband, let alone fibre, and you have consumers begging to pay for it. Who's going to foot the bill for redeveloping all our roads? Musk certainly wont.
> Things like this move slowly - it's not going to be retrofitted, but agreeing on a standard and new vehicles can implement in gradually.
The problem is you need it implemented on every vehicle for it to work. It's like anti-vaxers argument that "I don't need to get vaccinated when everyone else is" but it actually doesn't need to take a large percentage of unvaxinated people for an outbreak to still occur. Equally you only need one vehicle without a mesh network installed to case an accident that could kill several people - including those in autonomous vehicles. If you don't have total coverage then the mesh network isn't going to be reliable.