> The problem is the customers who are impressed by the human-sounding bot (sounding human is exactly what an LLM is for) and mentally ascribe human skills and

> The problem is the customers who are impressed by the human-sounding bot (sounding human is exactly what an LLM is for) and mentally ascribe human skills and thought processes to it. This is the key issue, I think. > it was never intended to be able to do this But to be fair, plenty of technologies start off that way until someone finds a way to make the technology do something it was never intended to do