Today, we’re launching a miniseries on underscoped problems related to catastrophic AI risks. For each area, we lay out what we understand, where the gaps are, and what kind of work would need to happen for the field to make progress. This writing is more speculative than our other work; think of these posts as public research notes rather than stakes in the ground.
What these areas need most are people Nan Ransohoff calls “general managers”:
“[General Managers] are founder-types who feel personally responsible for delivering a specific outcome (vs field-building generally); hyper-competent leaders who will pull whatever levers necessary to achieve the defined outcome. Most companies wouldn’t let an important initiative go unmanned or without a ‘directly responsible individual’ — why are we OK not having GMs for even more wide-reaching problems?”
Nan Ransohoff’s full post is worth reading. One of the examples she gives is D.A. Henderson, who led the WHO's smallpox eradication campaign. Henderson didn’t just focus on the industrial scale-up of vaccine production— that simply would not have been enough to eradicate smallpox. He tested deployment methods, flew to countries to see what was actually happening on the ground, and worked with governments to embed vaccination into national policy. He took full ownership of the problem from end-to-end, identifying the subproblems and making sure his research translated to real-world outcomes.
Many underscoped problems in AI risk call for that same mindset. We expect that careful, ambitious thinking about what to target will be as important as great execution. Some of these areas may turn out to be dead ends, whether because they’re ill-conceived in ways we can’t yet see, or because they require institutional buy-in that makes them unsuitable for independent researchers. General managers for problems in AI risk reduction should be prepared to face confusion head-on and to pivot quickly as the facts change.
Watch this space for our upcoming posts. If you want to own one of these areas, apply to the “Blind Spots” track of our CDTF program using this link.