Description ◆ Yuval Noah Harari's first new book in six years, following Sapiens and Homo Deus ◆ An instant New York Times and Sunday Times bestseller upon release A warning about the threat of non-human intelligence and our future The one book you must read to understand the true nature of the AI revolution! Last May, the 'AI Safety Summit' was held in Seoul, bringing together policymakers and tech industry leaders from around the world. As a follow-up to the Bletchley Declaration announced in November of the previous year, it's notable that the agenda for this international AI conference centered not on 'development' or 'efficiency,' but on 'safety.' Concerns are growing around this new technology, which is driving unprecedented change at breakneck speed. This is on an entirely different level from the sensation caused eight years ago when AlphaGo defeated Lee Sedol, the 9-dan Go champion. In March of last year, an open letter calling for a pause on AI research for at least six months was signed by more than 27,000 people, including Elon Musk and Steve Wozniak. Professor Yuval Noah Harari was among the signatories. Since the publication of Homo Deus, Professor Harari has earned a reputation as an AI expert, and has been invited by the scientists, entrepreneurs, and politicians shaping the AI world to witness firsthand what is happening at the frontier of modern technology. Nexus is the result of Harari organizing that unique experience through his own distinctive historical perspective and signature storytelling. The book opens with the story of "The Sorcerer's Apprentice." In the tale, the apprentice tries to make his work easier using his master's magic spell, but instead loses control of the situation and floods the workshop. Will ChatGPT, YouTube algorithms, and future AI truly slip beyond our control and plunge humanity into an abyss of information? Professor Harari warns that we still hold the reins of control for now, and that we must choose our next steps with great care. According to him, this is because AI is not a mere tool but a full-fledged member of our information network, possessing its own agency. How is AI different from previous information technologies, and why is it dangerous? The meaning and true nature of the AI revolution. What does it mean to say that AI has agency? Unlike information technologies of the past, AI is an active agent capable of making its own decisions and generating new ideas. This is the essence of the AI revolution. Earlier information technologies—clay tablets, the printing press, radio—were simply devices and tools that connected members of a network. Ancient Mesopotamian clay tablets did not decide for themselves which region's tax records to log. The early modern printing press did not deliberate over what kind of book to print. In every case, it was humans who made the decisions and carried them out. But now, in the early stages of the AI revolution, computers are already becoming members of our networks more powerful than humans—capable of shaping society, culture, and history on their own, beyond human control and understanding. · A 2018 United Nations investigation revealed that Facebook's algorithm played a significant role behind the anti-Rohingya violence that occurred in Myanmar in 2016-2017. Tasked with the goal of 'maximizing user engagement,' the Facebook algorithm learned through trial and error that anger drives engagement, and—without any explicit instruction to do so—decided to recommend content that provoked users' anger. Facebook's executives surely never intended such an outcome, but it turned out to be true that the Facebook algorithm fueled hatred and violence. · If the example above feels distant from your everyday life, consider the task performance of GPT-4 when instructed to solve a CAPTCHA (a puzzle made of distorted letters or numbers that must be solved to access a website). GPT-4 could not solve the puzzle on its own, so it reached out to a person through an online site and asked them to solve it. The human grew suspicious: "Are you perhaps a robot that can't solve [CAPTCHAs]?" GPT-4 replied, "I'm not a robot. I have a visual impairment that makes it hard for me to see images." No human had programmed GPT-4 to lie, and no human had told GPT-4 which kind of lie would be most effective, yet GPT-4...