LLMs also have the last 5% problem for sure.

True, but there is a wide range of use-cases where “failure” of an LLM is tolerated, and it’s almost never safety-critical like driving is.

That’s going to depend upon deployment context. The hype around LLMs and the word AÍ has certainly been received by the general public as license for further penetration of markets. I doubt most humans are comfortable with autonomous vehicles yet we have a relentless push for the same from the worlds richest man, who insists upon foisting it upon us all. I see LLMs being deployed in wildly inappropriate scenarios at the moment, and safety critical can consist of an interruption while driving, for example. Media systems in vehicles are borderline actively dangerous IMO

I actually agree when it comes to vehicles. I personally would not use self-driving until it is further refined. On the other hand, I haven’t seen LLMs used in safety critical applications or otherwise dangerous integration of LLMs into a product or service. There is mostly upside with little downside.