Strategy Fellowship
Applications have now closed for the MT24 iteration. We hope to run the strategy fellowship again in HT25 - sign up to our mailing list to keep updated!
Over 6 weeks, we’ll discuss key strategic parameters to help you develop your approach to AI risk.
- A 6 week reading group, from Week 3 to 8 of Michaelmas ‘24
- 2 hours commitment per week
- Exploring timelines, scaling, institutions and technical approaches in AI safety
Target audience:
- Those currently working on technical AI safety or AI governance, who want to understand how their work interacts with timelines and other strategic parameters
- Students and researchers who have only a little familiarity with the field of AI safety and want to understand the various approaches in a principled way
Content (subject to change):
- Timelines and scaling: the scaling hypothesis; biological anchors, the direct approach; trends in compute; takeoff models
- Threat models: what are the labs trying to build?; takeover; the difficulty of alignment
- The state of play: the chip supply chain; the labs’ plans; the role of researchers
- Partisan approaches: Situational Awareness; the difference between the US and China
- Slowing and containing: plausibility of slowing; race dynamics