www.lesswrong.com/posts/xJWBofhLQjf3KmRgg/four-ways-learning-econ-makes-people-d...
1 correction found
And here’s Daron Acemoglu assuming without any discussion that in the next 10 yrs, “AI” will not include any new yet-to-be-developed techniques that go way beyond today’s LLMs.
Acemoglu’s paper does discuss more-transformative future AI, including AGI, and explicitly says its estimates are conditioned on a particular task-level framework. So it is inaccurate to say this assumption is made 'without any discussion.'
Full reasoning
The cited paper does not simply smuggle in a 'no beyond-LLMs' assumption without discussion.
What the paper actually does is:
- Explicitly acknowledge AGI / much more transformative AI as a possibility. In the introduction, Acemoglu writes that some experts think 'artificial general intelligence (AGI) enabling AI to perform essentially all human tasks, could be around the corner.' That is direct discussion of AI going far beyond today’s systems.
- State a scope condition for the model. The paper says its macro estimates apply so long as AI’s effects are driven by task-level cost savings / productivity improvements. That is a modeling assumption stated up front, not an unstated omission.
- Say the paper is not limited to today’s LLMs. A caveat in the paper explains that although recent advances are in generative AI, the framework is meant to apply to 'other types of AI' as well.
So a fair criticism would be that Acemoglu chooses a framework that may understate discontinuous or AGI-like possibilities. But saying he made that assumption 'without any discussion' is contradicted by the paper itself.
3 sources
- The Simple Macroeconomics of AI (MIT PDF)
Some experts believe that truly transformative implications, including artificial general intelligence (AGI) enabling AI to perform essentially all human tasks, could be around the corner.
- The Simple Macroeconomics of AI (MIT PDF)
Although most recent advances are in generative artificial intelligence, the economic forces explored here apply to other types of AI.
- The Simple Macroeconomics of AI | NBER
So long as AI's microeconomic effects are driven by cost savings/productivity improvements at the task level, its macroeconomic consequences will be given by a version of Hulten's theorem.