The article “The Current Hype Cycle in Artificial Intelligence” explains why level 5 autonomous driving is an inflated expectation. My only disagreement is that it uses the flawed terms “artificial intelligence”, “machine learning”, “neural network”, etc.
If we called “artificial intelligence” with its real name, “probability modelling”, no-one would think that it is a limitless, mystical force, able to supersede humans and solve any problem.
If you call the basis of driver assistance technologies “artificial intelligence”, inevitably you are going to ask: “When will cars be able to drive themselves?” But if you call it “probability modelling”, the question is very different: “Could this lead to cars that drive themselves?” We would then have a healthy debate between yes and no; not estimates about 2025 vs. 2035.
Bad names often create a huge legacy of thinking the wrong thing. This is why we should stop using “autonomous” and “driverless” and use “driver assistance” instead.