Part 4/12:
While language understanding posed a substantial challenge, Dan believed the true difficulty lay in encoding common sense and comprehensive world knowledge into AI systems. Humans learn through experience and intuition, effortlessly bridging language ambiguity and understanding context—a feat that no system had yet mastered at scale.
He recalls being optimistic that scaling data and computational power would eventually bridge this gap but admits skepticism about ever truly infusing AI with the kind of innate commonsense reasoning humans possess. Decades of research had aimed to infuse AI with this understanding, but a truly comprehensive approach remained elusive.