One lesson connected to ChatGPT is that reliable pple aren't those who have good answers in general, but those who answer ONLY things they know v. well.
When lost, you'd rather get no direction than a wrong one.
Real doctors let you know what is outside their expertise.
@nntaleb Yes, they mention the limits of their expertise, but a good doctor (and scientist or engineer) is well-rounded enough to make reasonable estimates of things beyond their expertise and have a path to finding that out. The real lesson of ChatGPT is not to enshrine expertise-worship even further but to distrust even supposed experts. “Science is the belief in the ignorance of experts.” Everything an expert can actually *know* must be explainable on some level. Experts are fallible.
@Robotbeat @nntaleb good point, but got is it related to ChatGPT? Is it supposed to be the expert? Why?
@Robotbeat @nntaleb well, you're comparing failures in expertise with pretence for experience. I would avoid mixing the two https://www.theatlantic.com/technology/archive/2023/01/chatgpt-ai-language-human-computer-grammar-logic/672902/
@mapto @nntaleb People do that all the time, including experts. (The article makes a mistake by using the misunderstanding of the number of letters in a word as a problem of sentience… the model is actually totally blind to letters, it can only see tokens, ie full words or portions of words. If it ever gets the question right it’s only because it was able to infer the number of letters from other stuff in its training set, actually kind of impressive.)
@mapto @nntaleb My point is that the fallibility people see so easily in ChatGPT applies to all humans as well, and it’s a difference of degree, NOT a difference in kind…