Multi-agent intelligence explosions compared to human intelligence progress (evolution, culture, information)

Visualize empirical estimates of natural evolution, intelligence evolution, cultural evolution, and information evolution, compared to a simple multi-agent adversarial learning scenario, such as OpenAI's (2021) hide and seek, TripleSumo's (2022) cooperative adversarial learning, or any similarly complex environment. Use signals of 1) intelligence, 2) performance, and 3) power. Possibly relate it to military power based on embodied abilities of simulated robots (Rudin et al., 2022) to make it scarier (though be aware of dual-use). See Epoch AI's work and Hendrycks (2023) along with OurWorldInData.