Project summary

I'm concerned that AI time horizon's tendency to infinity and the way in which AI utility scales in time horizon are being poorly modeled, and that this poor modeling is leading to too little consideration being given to the possibility that ASI will arrive soon (i.e., within months rather than years).

I will create a quantitative model of AI progress which models these better to forecast intelligence takeoff. When it will happen and what it will look like. This will involve some careful reasoning about what it means for a model to have an infinite time horizon in a domain and what progress looks like beyond that point.

Here's a prototype model: https://jimfund.com/2026-vi.html

Importance

Helps with predicting intelligence takeoff which is helpful for AI safety decision making.

How will this funding be used?

To cover living expenses while I work on this project.