As AI advances, it will increasingly exceed our individual capacity to manage it. The counterweight is human collective intelligence.
Much of the discussion about AI risk focuses on misalignment, systems that deceive us or pursue goals we did not intend.
On the current trajectory, even AI that works exactly as intended will outpace our ability to understand it.
Systems can already process information and make decisions beyond what any individual can follow. As they become more capable and more widely available, that gap will widen.
Our authority over these systems is limited by our ability to understand them, and no individual will be able to keep pace.
However, human collective intelligence allows us to combine knowledge, experience and judgement to address problems beyond any one person's abilities.
Collective intelligence does not depend on any one person understanding the whole problem. It depends on people being able to question and build on what others know. AI could strengthen that process, allowing it to work more widely and more quickly.
Or it could bypass that process entirely, acting in ways that leave human understanding behind. Which happens depends on how we choose to design and use it.
We should be as ambitious about developing human collective intelligence as we are about advancing artificial intelligence.