After more than 10 years developing a self-driving system for cars, and having invested many tens of billions in development work, prototyping, and driving millions of miles with cars, leading companies in this field ( Waymo and Cruise ) they are still running very localized taxi services in some cities, and with very close supervision by remote controllers to solve possible problems.
In summary, it can be said that the companies have made an enormous effort, both human and economic, with still quite limited results.
And there are many other companies around the world also working in this field that remain behind them. So autonomous driving through cities and highways at the moment seems quite far from reality.
But instead, new companies have emerged that are trying to gain a foothold in this market, and have a different approach when it comes to developing the driving system. They do it based on Artificial Intelligence
They call this new approach: AV2.0 and it is based on the reinforced learning that has provided so many successes to AI to compete in various types of games (Go and Chess among others), and exceed expectations starting from scratch and providing just one series of “norms” to the machine.
There are several companies following this path (Waive, Ghost, Waabi, Autobrains, etc…) hoping that using smarter technologies they can catch up and even surpass the current leading companies.
These new AV2.0 companies do not use HD (High Definition) maps to help their cars, but rather rely on the car's GPS and sensors, and the AI with all the data it receives, learns to interpret the roads. I say it in a simple way, but the computer system on which they are based is very complex.
According to them, it will cost them a little more to start at the beginning, but later they hope to evolve easily from one city to another, because their system learns by itself.
It is true that when a company begins to develop a product using a certain route, then it is difficult for them to change, because we are all a bit of a creature of habit.
But I find it hard to believe that with the know-how that Google has on the subject of AI, it has not seen that this new path could shorten its autonomous driving developments. I am convinced that they are not fools.
These new companies are currently simulating their developments using the AV2.0 path, and they are still quite far from putting cars on the road, which is when they come into contact with the harsh reality.
Time, as always, will be the judge that shows us which was the most appropriate path. But of course I wish all of them the best.
EN 
ES