Discussion about this post

User's avatar
Kaleberg's avatar

Brink Lindsey almost had me until "It’s the arrogant and narrow-minded prejudice of scientism to suggest that this know-how is an inferior form of knowledge." Sorry, but it is. It can interpolate, sometimes successfully, but it can't handle novel situations. It can't make useful predictions outside of a narrow well explored space. It can't manage an internal model or maintain consistency.

Kaleberg's avatar

ChatGPT is a fascinating rhetorical device that lets us appreciate amazing usefulness of the Bayesian reliability of classical logic.

P.S. The ability of ChatGPT to pass the bar exam partially validates the criticisms of standardized testing. It is definitely possible to get good grades and pass tests by Bayesian mimickry. This suggests rethinking the tests, not eliminating them or suggesting that passing the test qualifies ChatGPT.

6 more comments...

No posts

Ready for more?