His point is that the remaining 5% of a problem can take considerably longer to solve than might initially be anticipated by looking at the quick progress made on the first 95% of it. A textbook example of that is voice recognition.

It's also not particularly clear that the costs will drop rapidly to zero or even to a level that leads to mass adoption (see air travel at Mach 1+, Segways...).

Your examples aren't very good, sorry. We're talking about adding AI, autopilot, autonomous capabilities, whatever you want to call it to already existing hardware. It's mostly a matter of software and you're a living example that it can be done, in 20W, with nothing more than stereo cameras and mics.

If I had to compare, I expect the change to be more akin to what Excel has done to paper spreadsheets.

Current voice processing tech is still very much worse than an average human, no matter how fancy the microphones and how much power it's allowed to consume from the grid.

Good sensors aren't cheap, the optical and mechanical elements aren't likely to drop in price soon. Only the computing power is getting cheaper, but harnessing it isn't particularly easy (we're not getting better at programming as quickly as we're making faster chips).

If you need another example, consider automated cleaning robots. It's a considerably simpler problem to solve, and they're still only a marginal product. At the moment, autonomous tech is only really used in well-controlled industrial environments. It will surely break out of there progressively, but chances are it will take a while.

we don't disagree in anything. i'm just saying that it will happen, it's a matter of time - 5 years, 10 years, maybe 20 years, but it's going to happen. there's thousands of smart people working on it, millions if not billions already invested and tens of billions at stake.

You're oversimplifying it and being a little pessimistic. Why? Because there's an order of magnitude or more difference in the complexity of AI of cars and AI that a human uses to navigate

That people are talking as if it's just round the corner, that taxi drivers and truckers are about to become obsolete. Whereas I understand the technical challenges are much greater than is stated, it may even be they won't be all overcome.

It's not that taxi drivers and truckers are going to become obsolete tomorrow. It's that that happening is such a huge cultural and economic change that we need to start planning for it today if we're going to not destroy the livelihood of millions of people when those industries do become obsolete in five or ten years.

About 20 million jobs disappeared in America last year, although there was a net gain in jobs overall. This is going on all the time.

So if all of America's 11 million professional driving jobs were to disappear over the span of a decade, it would be a blip, not an apocalypse.

Tesla is launching Autopilot 2.0 (in stages) starting in January.

It's still NHTSA level 3, not 4 or 5, but it's a major step up from pretty much anything currently available and will be in consumer's hands. Even more so when the Model 3 comes.

I think you'll be surprised how fast this technology will exit the labs. Just that Google has decided to brand around it is a big indicator that they're growing confident.

But so what? Full autonomy needn't even be a business goal of Waymo. They just want full deployment. Having a system in place which reduces or eliminates many/most of the accident root causes would be worthwhile to install on any vehicle on the road even if it technically had a human driver.

The "goal", such as it is, is to have a technology worth using. It's not to transcribe whichever comic book utopia someone has in mind.

Their stated goal is indeed full autonomy.

The point is that we are far far further away than commonly accepted. The day a computer can distinguish between a plastic bag in the road and a child is the day I'd support them on the road.

Do computer drivers have to be perfect, or is it enough to be much better than human drivers?

Perfection is of course, impossible, the best we can hope for is progress asymptotically close to ideal. The minimum bar should be performance surpassing human performance, but the details do matter. For example, a 99.99% safety rated car which almost never crashes but randomly explodes once every 10,000,000 uses would probably not be accepted.

Or if you see a ball cross the road there are strong odds a child might be chasing after it.

The day a human can distinguish between a plastic bag in the road and a child is the day I'd support them on the road.

My mother once almost killed a child who ran into the road. The kid was on the side of the road and suddenly ran into the street. She did not see him, and he ran into some sharp piece of metal on the car (this was in the early 60s). He bled profusely, and the neighborhood people came out and almost became violent with her, blaming her for the incident.

If she had been in an automated car, it would have almost certainly have stopped before the collision.

And what's wrong with slowing/stopping for a plastic bag? How often do plastic bags really sit in the street shaped like babies?