When Machines Forget Better Than Humans

1.    In the near future, AI might not just assist us but it might outlearn and out-unlearn us. Machine unlearning is already evolving with techniques like SISA, Approximate Fisher Forgetting, and Influence-Function–based Unlearning, allowing models to selectively forget data. While still imperfect, AI’s ability to “forget” deliberately and at scale could soon surpass human capability.

2.    Humans, by contrast, struggle with unlearning. Our beliefs, education, culture, and even genetics shape what we retain sometimes stubbornly. A simple fact, like Pluto’s planetary status, shows how knowledge once unquestioned can become obsolete, yet unlearning it completely is never easy.

3.    As technology accelerates, generational gaps widen, and obsolescence hits faster, the gap may grow: AI may unlearn and adapt faster than we can. This raises a provocative scenario: will humans remain HITL (Human-in-the-Loop), or transition to AIITL (AI-in-the-Loop)—where AI doesn’t just assist but guides, corrects, and even reshapes human understanding?

4.    Unlearning, whether human or machine, is imperfect. But in the coming era, recognizing the limits of our own memory—and the power of AI to surpass it—may become humanity’s most urgent lesson.