How to think about AI: The Economic Model

Arguably, present AI systems as well as those on the horizon don’t fundamentally offer new capabilities. They make easier things that could eventually get done by a human–or larger group of humans. (If you have counter-examples, please leave them in the comments below).

For example, a code-completion LLM like Github Copilot does what an new-grad with an unusually deep history of looking at code could conceivably offer. Diffusion model capabilities pretty much align with commissionable art tasks (or fall short–depicting hands still require professional art skills to get right).

And so on, with countless more examples.

Now, to be clear, the quantity of output possible produces its own quality. It’s not like there’s been a history of people with money to burn bringing on new-grads to study code for years to finally be their personal coding assistants, or commissioning a dozen original art pieces daily to put in their newsletter.

On the whole, AI presents amazing new capabilities. Careful viewing through an economic lens can help temper some of the hysteria associated with the hype-cycle we’re in. It can suggest new possibilities. At the same time, economic issues raised by recent AI advances cause their own problems…but that’s for another article.

Originally posted on LinkedIn. 100% free-range human written.