How expensive something is is a very poor starting point for what we should attempt to automate with nascent technology that still has a pretty high miss rate

How expensive something is is a very poor starting point for what we should attempt to automate with nascent technology that still has a pretty high miss rate It’s not entirely irrelevant to the commercial case but you really need to start with technical feasibility and how resilient the job is to mistakes. And something that is high risk, high leverage and very softskills/experience driven is a bad place for AI imo Anthropic‘s shenanigans with their AI vendor should illustrate how far we are from credible AI CEOs