Having listened to many hours of Runciman’s podcasts (formerly Talking Politics, now the also-excellent Past, Present, Future), it’s always a treat seeing how well his voice translates into his writing. He’s approachable, inquisitive, and authoritative in equal parts, all of which makes it much easier to follow him along his more tenuous tangents.

The Handover is a more thorough exploration of a thread he’s been pursuing since at least How Democracy Ends, namely the conception of states and corporations as “artificial agents,” essentially the decision-making equivalents of “artificial intelligence.” It’s a helpful framing for understanding just how much agency we’ve already handed off to non-human entities, and why it seemed like a good idea to essentially farm out many of our key responsibilities to institutions with no intelligence and no humanity. The issue, then, is whether AI will be a tool for humans, or a partner for our artificial agents that ultimately aligns against us—not in a paperclip machine sort of way or anything quite so dramatic, just in a way that further alienates human interests from how the world runs.

The most useful thing about the “artificial agent” framing is the reminder that there’s a big difference between “novel” and “unprecedented.” AI is new, but we’ve been handing off responsibility to “machines” for centuries, and Runciman reminds us that there’s always a historical context worth learning from.