An interactive model of AI takeoff speeds

Jaime Sevilla and Eduardo Infante Roldán
We have developed an interactive website for understanding a new model of AI takeoff speeds.
Website

Tom Davidson from Open Philanthropy has released What a compute-centric framework says about AI takeoff speeds, a report investigating how fast AI capabilities might transform the economy.

Epoch has supported this project by coding the model and running the simulation experiments required for the investigation. As a supplement to the report, we have developed an interactive website presenting the model and some of the report’s results.

This website includes several sections and features, which we briefly describe below.

Playground

In the playground, you will find an interface to enter the parameters of the Full Takeoff Model, and see how these affect the results. It includes graphs of the most important variables of the model, as well as tables summarising the results.

Reports

In this section we show four reports:

  • The Monte Carlo analysis shows a summary of 10k samples of the model’s parameter values.
  • The aggressive Monte Carlo analysis is the same, but using a more aggressive distribution for the amount of 2022 FLOP required to automate all productive tasks.
  • The parameter importance analysis shows the results of a superficial sensitivity analysis of the parameters of the model.
  • The timelines analysis is a side-to-side comparison of several model scenarios.

The Monte Carlo analysis is specially important, as it summarizes quite well the AI development scenarios that are most plausible under this model.

Figure 1: Graphs showcasing the Monte Carlo distribution of takeoff duration and 100% automation date.

Description

Here you will find an interactive mathematical description we wrote of the model and all its nuances.

We think this section will be most useful for readers with a background in economics or mathematics who want to understand how the model is built.


We are actively developing the model and researching extensions. We invite you to contact us or open an issue in the associated GitHub repository if you have comments. Please speak with us if you’re interested in our work and have a background in optimising large Monte Carlo analyses.

We hope you will find our website a useful tool for understanding Tom Davidson’s report, developing your intuitions about the development of AI and for future research!


If you want to contribute to our research, consider filling our expression of interest form.