What self-driving cars really need is an express lane where there are no surprises and a constant signal from special markers. On their own turf, they could run cars and pallets of freight at near-zero clearance. You could request a merge into the next passing gap, punch in your exit number, and watch screens until you got close to it. Also, there is no good technical excuse for any land vehicle to weigh more than it carries. We are accustomed to absurd waste because it generates profits.

I am excited for this. I think people don't understand just how much faster technology improves today than it did even a few years ago.

FSD customer beta testing as Tesla does should be stopped because it evolves all the people around the car and none of those agreed to the risk. This use of the general public as test subjects without consent is just wrong.

There driving tragedies that happen without fault, unavoidable. We drive anyway. We choose to take the risk. FSD may prevent 90-99% of this. It may be 100-1000 times safer. And still, some people will say it's too dangerous and shouldn't be allowed. This is an irrational fear of tech. It's nothing new. But this is what FSD has to overcome.

So, Autopilot drives easy parts of journey, hands back control to driver the moment things get difficult, and still is only slightly better than driver? Comparing with average is comparing with reckless drivers that wouldn't be able to afford expensive new car. Curious how comparison would look like with similarly priced cars, ideally driven by similar drivers.

What is really needed to make this a usable reality is Vehicle to Vehicle (V2V) communication and Vehicle to Road (V2R) communication. Once vehicles can talk to each other, the whole system works much better. When you add in the ability of talking to the road, you get a system that is nearly completely safe. The road needs to be smart and able to see just like the vehicle does. You can make it anonymous by hiding identifying details although I'm sure that some people will fear the "big brother" aspect of streets being able to communicate about pedestrian and other non-automobile traffic. V2V combined with V2R lets every vehicle see things that it can't see visually. If, for example, the car need to make a left on to a crowded street, V2R lets vehicles report their location and speed to the road. The road, in turn, can report this data to the car wanting to turn. The road can "see" what the car cannot.

Self driving doesn't need to be perfect.
Perfect is really hard.
It just has to be better than humans.
Better than humans is really easy.

I look forward to lower insurance rates and seeing lower numbers of idiots dying on the roads when the auto pilots take over for knowingly distracted drivers. Looking at the last video, I was amazed at how many fools were halfway in the road with cars coming every ten feet on either side. That one lady had her big woven bag 2 feet into the street, like she was daring someone to hit it. I guess the AI is progressing pretty well. (Just don't give one to Cyberdyne Systems for their Skynet work.)

Correct me if I am wrong, but wasn't a kid recently run over by a self driving car? And how about hackers, I know for a fact that computers in cars have been hacked while being driven, some even having their engines shut off rendering power steering inoperable. I do not trust machines to do my driving

The auto insurance industry will get on board when these vehicles can do better than the average human driver. The current claims rate is just under six per hundred vehicle years, each of which represents 13,500 miles on average. The average collision claim was $3500. Liability claims were a quarter as frequent but each event was four times more costly.

Across all drivers this is when claims costs less than $7000 for a quarter million miles driven. If you look at age groups the under 30 and over 75 groups have significantly higher claims cost, so they may be insurance industry's first targets. Persuading people that a car can drive better than they can might be a challenge.)