Pretty sure that is Chet from Weird Science.
The first to market doesn’t always win in the long run. Competitors have a positional advantage. They have information about what the first guy did and can introduce improvements. The Wright brothers didn’t win aviation. Edison didn’t win electrical power. In most cases nobody even remembers the first guy. I wouldn’t bet on it but Tesla could come back. Chances better if/after Musk is shitcanned.
FSD = “Fully Self-Driving”?
2 Likes
Didn’t he though? My understanding is that he pitched alternating current vs. Tesla’s (the person) direct current and won out, maybe thanks to the Chicago World’s Fair. Or at least he was given the contract to light the fair, something like that.
You’re backwards. Tesla advocated AC. Edison was DC.
4 Likes
Hand grenade 
Edison, among other things, staged fake animal deaths (that he murdered) to protect DC.
It was a stroke of luck AC won out when it did. Something to do with Niagara Falls.
2 Likes
It was inevitable that AC was going to win out, if you’re just talking about timing then maybe, I don’t know. But any large scale electrical power distribution system has to be AC.
Or rather, had to be AC. With modern electronics it’s actually possible to design a practical large scale DC power system, but that wasn’t practical back in olden times. And I think that there are some very long distance high voltage DC transmission lines now, since high voltage DC transmission is more efficient than AC over long distances.
1 Like
I think it really just comes down to what you are telling the AI to optimize for, and then it’s just going to do its best based on its learning experience.
I know nearly nothing about machine learning, so maybe I’m totally wrong.
5 Likes
It really comes down to when we commit to roads optimized for AIs. Maybe the point of the Boring Company is to create roads where self-driving Teslas can’t fuck up.
1 Like
Given regulatory realities and the average age of the car fleet of 12 years, 30-50 years away from true self driving cars being close to ubiquitous in an off highway environment in the US
Sure seems like a better idea to optimize the cities for humans than to optimize roads for AI cars. Or I guess we can keep on keeping on with all the car-dominated shit that makes so much of our world such a horrible place to be. Build all the Torment Nexuses because fuck it, yeah, why not?
4 Likes
Probably true, especially if one cares at all about climate change. EV shift has its own problems and won’t finish close to fast enough to matter if climate models are close to correct. Need to move away from complete freedom to travel model.
I love everything about cars and car related infrastructure except for the ability to see various sights and environments while I travel, so this all sounds great for me.
1 Like
I mean, are they going to move all the mcdonalds and booger King’s underground as well? Because I could get used to that.
It’s not really freedom to travel that needs to be sacrificed. It’s freedom to live at stupidly inefficient densities. There are plenty of places on Earth where not owning a car is a practical choice.
2 Likes
A reasonable future state might be something like large, dense, walkable neighborhoods with car friendly routes in between that include AI cars. One of the solutions to the “AI cars can’t do simple things like stop running into pedestrians” is to stop having cars and pedestrians in the same places all the time. If you see a pedestrian on the highway you’d think that person is nuts and shouldn’t be there. Arguably we should all have the same reaction when we see cars zooming through walkable neighborhoods full of pedestrians. But somehow we’ve convinced ourselves the second situation is a necessary evil,
1 Like
You basically give modern AI a set of basic rules like don’t hit solid objects, the rules of the road and how to operate a car then provide it with a large dataset of people driving around the real world generating the inputs and outputs the AI uses and the AI will mimic that. In theory these days you don’t even need to give it the rules of the road as it would figure those out but hard rules like road rules are normally given. It will then eventually start behaving similar to those people in its training dataset. Then you let it lose in the world and it will continue to learn from all mistakes all cars with the same AI make. The actual internal model of the AI is no longer something we can fully visualise so we also visualise the inputs the AI gets and get it to identify them. This is more for humans to try and find what went wrong as not all accidents are a mistake by the AI. It could be that its lidar or camera never had the contrast needed to identify something us humans can which is not an AI issue.
Eventually the AI will drive better than us as it has access to more inputs like lidar, doesn’t need to move its head to look behind and has vastly superior reaction speeds.