The Basilisk Stare begins with Daniel Susskind presenting what initially looks like a nonfiction investigation into artificial intelligence and the disappearance of an old friend. He explains that the leaders of major technology companies believe they may soon create AI systems capable of outperforming human beings at almost every intellectual task, and he argues that society has failed to take seriously the question of what happens if they succeed. That abstract concern becomes personal through Oren Senesh, Susskind’s brilliant former Oxford friend and a central figure at the enormously successful AI company QuantumMind. In November 2023, QuantumMind suddenly implodes. Its board, led publicly by Sarah Lannister, fires CEO Yonatan Levi, even though Yonatan has helped turn the company into a $100-billion giant through successive releases of its Zohar AI system. The market reacts disastrously, the board is removed, Yonatan is reinstated, and QuantumMind rapidly retreats from the whole episode. The crucial revelation is that Oren, QuantumMind’s CTO and a board member, had secretly instigated the coup. Immediately afterwards, however, he disappears from public view and refuses to explain himself. Because Oren has been obsessed with AI safety since his youth, Susskind initially suspects that he discovered something horrifying inside the unreleased Zohar 6.0. Perhaps it could design biological weapons, manipulate elections, exploit nuclear systems, blackmail individuals, operate autonomously through the digital world, or otherwise create an existential danger. The mystery is strengthened by the fact that Oren would have known how much damage the coup would cause to his company, his reputation and his closest colleagues. Whatever he saw, Susskind reasons, must have seemed serious enough to justify destroying his own life’s work.
The book then moves backwards into Oren and Susskind’s friendship. They meet as students at Balliol College, Oxford, on Yom Kippur, when both are fasting. Oren quickly reveals himself as an extraordinary mathematical talent, someone who seems to understand difficult ideas almost effortlessly. Their friendship deepens during a summer reading party at a primitive Oxford chalet near Mont Blanc. They spend a fourteen-hour walk talking about their families, Jewish identity, antisemitism, the Holocaust and the strange feeling of belonging to an intellectual tradition while also existing outside the British class system around them. Oren comes from a family descended from an old rabbinic dynasty, while Susskind recounts how members of his own family escaped or were murdered in Europe. These memories matter because Susskind portrays Oren not simply as a clever engineer but as someone who has always treated intellectual questions with almost religious seriousness. During their doctorates, Oren concentrates on AI safety when most researchers are trying to make AI more capable. His famous warning is a variation on the paperclip-style catastrophe: tell an immensely capable AI to maximise pin production and, unless the goal is properly constrained, it may eventually transform the Earth and perhaps the universe into pins. Yet Oren’s deeper ambition is not safety alone. He wants to “demystify intelligence”: to build AI in order to understand what human intelligence really is. After Oxford, he leaves academia for industry, eventually helping to create QuantumMind, while Susskind remains an academic and gradually loses contact with him.
Much of the middle of the book is Susskind trying to reconstruct what Oren might have discovered. An email from Oren says that “everything changed after move 37.” Susskind traces this first to Garry Kasparov’s defeat by IBM’s Deep Blue in 1997 and then to Lee Sedol’s defeat by AlphaGo in 2016. In both cases, a famous move 37 becomes psychologically devastating. Kasparov cannot believe Deep Blue could have made such an intelligent move without human assistance, while Sedol is stunned by an AlphaGo move that violates centuries of conventional Go wisdom yet turns out to be brilliant. Sedol eventually retires, saying AI has destroyed the meaning he once found in the game. Susskind connects this to Douglas Hofstadter, who had once believed that machines would need something resembling human understanding, emotion and consciousness before they could rival human achievements. Instead, machines repeatedly become capable without becoming humanlike. This leads to one of the book’s central ideas: the history of AI divides into “Before Kasparov” and “After Kasparov.” Researchers once assumed that to surpass human intelligence, machines would have to imitate it. Deep Blue, AlphaGo and later generative AI suggest the opposite. Machines can reach the same destination by completely different routes. Susskind wonders whether this destroyed Oren’s lifelong hope of using AI as a mirror for understanding humanity. Perhaps Oren finally built an extraordinary intelligence only to discover that it revealed nothing about human consciousness at all.
