Audio version available at Cold Takes (or search Stitcher, Spotify, Google Podcasts, etc. for "Cold Takes Audio")

This piece starts with a summary of when we should expect transformative AI to be developed, based on the multiple angles covered previously in the series. I think this is useful, even if you've read all of the previous pieces, but if you'd like to skip it, click here.

I then address the question: "Why isn't there a robust expert consensus on this topic, and what does that mean for us?"

I estimate that there is more than a 10% chance we'll see transformative AI within 15 years (by 2036); a ~50% chance we'll see it within 40 years (by 2060); and a ~2/3 chance we'll see it this century (by 2100).

(By "transformative AI," I mean "AI powerful enough to bring us into a new, qualitatively different future." I specifically focus on what I'm calling PASTA: AI systems that can essentially automate all of the human activities needed to speed up scientific and technological advancement. I've argued that PASTA could be sufficient to make this the most important century, via the potential for a productivity explosion as well as risks from misaligned AI.)

This is my overall conclusion based on a number of technical reports approaching AI forecasting from different angles - many of them produced by Open Philanthropy over the past few years as we've tried to develop a thorough picture of transformative AI forecasting to inform our longtermist grantmaking.

Here's a one-table summary of the different angles on forecasting transformative AI that I've discussed, with links to more detailed discussion in previous posts as well as to underlying technical reports:

Forecasting angle	Key in-depth pieces (abbreviated titles)	My takeaways
Probability estimates for transformative AI
Expert survey. What do AI researchers expect?	Evidence from AI Experts	Expert survey implies¹ a ~20% probability by 2036; ~50% probability by 2060; ~70% probability by 2100. Slightly differently phrased questions (posed to a minority of respondents) have much later estimates.
Biological anchors fra

...

Intro to AI risk