Why AI Spending Isn't Slowing Down

As the theory goes, there is a network effect forming. Early AI products with strong adoption get the users to implicitly and explicitly label data. This leads to improvements that are a moat due to “data flywheel”

So there’s a “first mover” type advantage because you can’t get that usage data anywhere else - not even paying participants - because there exists a domain gap between offline and online measurement.

There is a reasonable counter argument of “the bitterest lesson” which says all that matters is the improvements in foundational models.