Artificial intelligence has been a stated goal of computer science for decades, and this Off Book episode asks why building it keeps turning out to be so hard. Ernest Davis of NYU's computer science department and psychologist Gary Marcus, also at NYU, lay out the case for skepticism: humans took millions of years to evolve the pattern recognition and common sense that let us navigate an absurdly complicated world, and no algorithm has matched that yet. Yann LeCun, of NYU's Center for Data Science, describes the current approach of modeling AI on the brain itself, simulating the mechanics that produce learning rather than hand-coding rules. Robin Hanson, from Oxford's Future of Humanity Institute, widens the frame to what happens if researchers succeed, from automated labor markets to machines that force a rethink of what intelligence even means. The episode moves between these four researchers rather than settling on one verdict, tracking real progress in language processing and image recognition alongside the honest admission that human-level machine intelligence is still, by most estimates, a long way off.