Wouldn't that objection disappear once automation is more reliable than a human operator, since the human operator would be less likely to be sued when taking responsibility for an automated system's results than they would for simply trying to make the call themselves?

It's more about who gets sued. If you make software to fully automate interpretation you would be the party sued. If there is still a human in the loop, you are not liable for their errors.

I don't understand how that works - if you're an employee of a company, generally the company is liable for mistakes you make (unless you're also a director of the company). So for the company, whether they choose to provide their services based on the output of some software or based on the judgement of a human employee, the result should be the same.

I can see an argument that if the company was sued then it could try to push the blame onto the software vendor, but surely that would be decided based on the contract between company and software vendor, which is usually defined by the software license.

The person operating the machinery is not working for the company that makes it though.

This seems absurd. Are there any examples you can point to, or legal principle you are referencing here?

But in many areas of medicine, solutions have existed that outperform humans for years (not related to current deep learning wave). Yet they were not implemented for regulatory/legal reasons.

Can you share any examples? I'm a non-medical human and curious.

This suggests a lucrative area of legal practice: suing for negligently failing to automate.