7 Comments
User's avatar
Radek's avatar

“And there is some evidence that they are filtering their data to get the results they expect”

That is not at all what that linked paper says

Joseph Francis's avatar

“Our findings underscore the critical importance of data cleaning in shaping applied microeconomic results.” It’s my interpretation of that. Possibly it’s the wrong interpretation?

Radek's avatar

That just says that how you clean your data affects results. There’s nothing there about “get the results they expect”

Joseph Francis's avatar

Yes, I am reading between the lines. But yours is an alternative interpretation: perhaps the effects of data cleaning are random and don’t bias the results in favour of the authors’ argument. I may be overly cynical.

Cian O'Connor's avatar

I've been thinking for a while that the problem in a number of areas is not that LLMs are able to outdo humans, but rather that we've created a society where humans are incentivized to behave like robots. And we've reached the end point of that - the robots are better robots than we could ever be.

PEG's avatar

Yep, the business version is desire for “robust processes”. It’s gone from supporting quality to removing all contextual human elements, which has resulted in a lot of the organisational pathologies we see today.

Oakie McDoakie's avatar

Because of the rise of AI and the growing importance of non-written mediums, I've been thinking for years that the focus should go back to teaching the trivium—grammar, logic, and rhetoric—with an emphasis on doing so orally, in class.