Hacker News Reader

14:17: "I suspect too much of the processing power is going into using the model as a database instead of using the model as a reasoning engine"

Can you (or someone) expand more on what he means by this?

asking and retrieving info instead of iterating through analysis steps with inputs toward some goal

When you ask GPT to add it probably tries to remember a similar instance in the training set rather than performing the calculation on the spot