“Chamath Palihapitiya predicts that recursive self-learning in AI models could lead to an order of magnitude improvement on a yearly basis, creating a new form of Moore's Law.”