The fundamental problem with alignment lies in AI inheriting conflicting and undesirable human values: AI may harm despite, and even because of, human-machine value alignment. To ensure AI benefits society, Ghebreab argues, a shift is necessary from human-machine alignment in AI, to human-human value learning and alignment through AI. According to the professor, this is the only way out of the current unhealthy catch-22 relation between society and AI.
AI has been likened to numerous metaphors, but to the professor the glass mirror stands out as one of the most compelling and telling, given the profound technological, philosophical and social associations that have evolved around it over centuries. Like the glass mirror, AI has the potential to transform societies for the better, over time and space. However, this potential is at risk of going unfulfilled, Ghebreab warns, because AI is currently being developed and deployed within self-serving and self-perpetuating societies, driven by the “ethics of the market”.
Ghebreab calls for courage to think out-of-the box and adopt a perspective on AI that emphasizes inclusive engagement with the planet and its rich diversity of people, cultures and values. In this perspective human-human value learning and alignment should take precedence over human-machine alignment. With this an alternative path is possible to more self-critical and self-conscious societies, based on the “ethics of life”.
Prof. dr. S. Ghebreab, professor of Socially Intelligent AI: Infinity, Kaleidoscopy and Society: On reflections of AI.
You can watch this inaugural lecture here.