Can artificial intelligence unlearn what it has learned?

The latest wave of artificial intelligence systems is facing a new challenge as legal and policy frameworks scrutinize their learning capabilities. The captivating question at the forefront is: Can AI forget what it has learned? And can the European Union’s “right to be forgotten” regulations also apply to AI?

In the midst of this debate, a classic Turkish art music song by Nesrin Sipahi sets the tone, emphasizing the unforgettable nature of memories. But should AI systems also possess this ability to remember indefinitely?

With AI already grappling with copyright issues, the concept of the “right to be forgotten” in search engines and social media has sparked discussions on its extension to AI. This rule allows individuals to request the removal of personal information from online platforms.

An article by Politico delves into the complexities of this issue, highlighting the challenges of making AI systems forget information they have learned. Researchers are exploring the concept of machine unlearning to develop methods that allow AI models to selectively forget or approximate their training data.

While erasing learned information from AI systems poses a significant challenge, researchers believe that finding easier ways for AI to forget could facilitate compliance with agreements and regulations. As regulators in Europe aim to apply existing digital laws to new platforms, the question of whether AI can forget what it has memorized about people remains a pressing issue in the realm of artificial intelligence.