Takeaways:
👉#AI returns saturate if physical lags
👉Wages often rise & fall as intelligence automates
👉Wage declines require shrinking intelligence-job share
👉Policy: pace deployment + invest in physical capacity; don’t bet on a “singularity”

Result 4 (simulation). Here we simulate adding more AI, during and after automation. (1) Wage effects of adding more AI saturate. (2) Higher substitutability between physical and intelligence can lead to wage declines during automation but allows for unbounded wage growth post-automation.

Result 3 (simulation):when intelligence and physical are more substitutable, automation of intelligence has a larger positive effect on output.
➡️What’s better for workers’ wages is worse for output, trade-offs.

Result 2 (simulation): when intelligence and physical are more substitutable, automation of intelligence has a more positive wage effect early on but a more negative effect later.
➡️Cautionary tale, early wage gains may not persist!

Result 1: as automation of intelligence by #AI progresses, output increases and there’s an increase in the share of workers in the physical sector \beta (i.e., in person work, including e.g. cooking, teaching and surgery).

Model: Output is produced with intelligence and physical sectors (CES), each powered by humans and capital (AI is the capital in the intelligence sector). Labor optimally reallocates between physical and intelligence to maximize wages.

Is intelligence saturation plausible?
(1) More intelligent people are not proportionately more successful.
(2) Adding more scientific researchers yields lower and lower additional output
(3) Prior tech (ICT) revolution had limited growth impacts.

Why distinguish physical and intelligence? Intelligence saturation: intelligence can make physical inputs maximally efficient, but the impact saturates.
➡️Physical and intelligence sectors are complements, you need more of both to increase output.

We've made an interactive tool you can play with to see how wages would evolve with automation & with adding exponentially more #AI, depending on parameters of your choice. Have fun (but maybe read the paper first 😈)! intelligencesaturation.org

Application link here: www.linkedin.com/jobs/view/43...

The conversation is based on my paper here: www.annualreviews.org/content/jour...

Some context: I used to ask “what are you working on?” but I learned some people would rather not talk about it for various reasons. So the more general “what are you excited about in your field?” lets them EITHER talk about what they’re working on or some other interest.