3.2 · How Much Should AI Actually Do?
Watch: Autonomy Gone Wrong
⏱ 8 minCourse 03
Zeynep Tufekci is a sociologist who studies the interaction between technology, society, and democracy. This talk is about how social media platforms use AI to maximise engagement — and how that optimisation objective, run autonomously at scale, produces outcomes nobody designed and nobody controls.
The Autonomy Link
The platforms Tufekci describes didn't make a single decision to radicalise users. They made a single decision to optimise for engagement — and then gave the AI Level 5 autonomy to pursue that objective billions of times per day. The harm emerged from the combination of the objective and the autonomy level. This is the lesson: you cannot separate the goal from the autonomy level.
