You begin with the concept or facts and then explain with a more understandable illustration or narrative. This technique is featured in a unique way in a new book on AI’s future.

Kai-Fu Lee’s most recent book AI 2041 offers a detailed look ahead at AI in the next 20 years. The book is cowritten by award-winning science-fiction author Chen Qiufan, and each of the 10 chapters begins with a short story by Qiufan, followed by an explanation of the AI technology at the heart of the narrative by Lee. Lee’s AI qualifications are extensive, having served as president of Google China; senior executive at Microsoft, SGI, and Apple; and cochair of the Global Artificial Intelligence Council at the World Economic Forum.

Lee calls the book a work of “scientific fiction,” a fusion of fiction and popular science as a way of exploring the future. The stories in each chapter are “portals” that lead you to the science. Lee began with a “technology map” that projected when particular forms of AI would mature. “Qiufan then dreamed up the characters, settings, and plotlines that would bring these themes to life. The stories cover all key aspects of AI from basic to advanced tech.” The first seven chapters examine tech applications for different industries, and the last three focus more on social and geopolitical issues.

Throughout the book, important themes are repeated and illustrated. One fundamental belief expressed early by Lee is that “AI is mankind’s final step in the journey to understanding ourselves.” On a practical level it’s an omni-use technology that he thinks will penetrate virtually all industries. And like most technologies, it’s inherently neither good nor evil; it’s one that he expects will eventually produce more positive than negative impacts on our society.

Lee frequently reminds the reader that “the ‘AI Brain’ (deep learning) works very differently from the human brain,” and he offers a detailed comparison of the strengths and weaknesses of human vs. AI “thinking.” He writes, “Deep learning’s decisions are based on complex equations with thousands of features and millions of parameters. Deep learning’s ‘reason’ is basically a thousand-dimensional equation, trained from large quantities of data. This ‘reason’ for producing a given output is too complex to explain fully to a human.”

In one discussion of Google’s GPT-3 natural language processor with its 1.75 trillion parameters for learning, he speculates, “Perhaps in 20 years GPT-3 will read every work ever written, watch every video ever produced and build its own model of the world. This all-knowing sequence transducer would contain all the accumulated knowledge of human history. All you have to do is ask it the right question.”

The explanations lead inevitably to the question “So, will deep learning eventually become ‘artificial general intelligence’ (AGI), matching human intelligence in every way?”

Lee’s direct answer repeats his insistence that human and AI “minds” are different. In fact, he suggests that we stop using AGI as the ultimate test of AI development. “What’s important,” he explains, “is that we develop useful applications suitable for AI and seek to find human-AI symbiosis rather than obsess about whether or when deep learning AI will become AGI. I consider the obsession with AGI to be a narcissistic human tendency to view ourselves as the gold standard.”

Chapter four is one of the most interesting in the book because its subject, healthcare, is the industry Lee sees as the one most likely to be transformed by AI. The confluence of digital healthcare has already begun. Radiology has recently become digital with computer visualization of high-definition 3D imagery. Wearable devices continuously monitor critical functions, robotic surgeries are commonplace and will continue to develop, and precision medicine, tailoring an individualized treatment for patients, is becoming more practical with the help of AI applications. And research for new drug therapies now has a great new tool for its second step of drug discovery—finding the 3D structure of a disease’s protein sequence (protein folding). In 2020, Google’s DeepMind developed AlphaFold 2 for this task, and Lee describes this as “AI’s greatest achievement for science to date.”

The stories are interesting, and they provide more than just a secondary perspective on the AI issues. Like Plato’s parable on how humans know what they know, the stories offer a more “familiar” view of Lee’s vision of the future of the technology that many think will define us and our world in more ways than we can yet imagine.

About the Authors