In 2020, a very down year, the Bureau of Transportation stats[1] give 1,934,743 MILLION miles driven. This Verge article[2] shows for 2020 autonomous vehicles drove 1.99 million miles in California. Let's be generous and assume across the rest of the US it's equal to double that, given Cali is the hotbed of testing at the moment. That puts us up to basically 6 million miles driven.
You're attempting to compare:
1,934,743,000,000 to
6,000,000
The data is not in yet on current generation autonomous vehicles.[1]: https://www.bts.gov/content/us-vehicle-miles
[2]: https://www.theverge.com/2021/2/11/22276851/california-self-...
The one point where I agree is there aren't as many of them as there are human cars, and it's possible that in large numbers there could be some unintended consequences. Think 1000 of them getting stuck at the same location. There should be provisions limiting how fast they can scale up to make sure that doesn't happen.
No there isn't. Driverless cars drive under a fraction of the conditions and in self selected locations precisely because the tech is still shoddy. Not to mention they only function at all because they're vastly outnumbered by humans who know how to respond to them.
Let's do an actual experiment to compare. 100% driverless cars with current tech, from different companies on all road conditions that humans drive on during busy hours in a city and see how that goes. To even attempt to compare driverless cars to human drivers without seeing the dynamics of a non-trivial amount of them interacting, which humans need to do all the time, is meaningless.