The Age of AI: And Our Human Future Paperback – November 1, 2022

What is this book? A treaty? A compendium? An impassioned plea? Or a sales pitch drawing its unique proposition from philosophy, history, and humanities littered with revelations of opportunities and doom?It is everything, and more.As of writing, GPT-4 is the rave in artificial intelligence (AI). But the world has already had intimations of AI's power through its predecessors, particularly ChatGPT. In a deluge of information, it is necessary to have a voice of authority and wisdom to explain the phenomenon we face. This book fits the bill as a pamphlet that conceptually explains what AI is, without the technicalities that may baffle non-technical readers. (I have enjoyed an interview between one of the authors, Eric Schmidt and Alexandr Wang, on the subject, which inspired my purchase of the book).As a compendium, the book catalogues the development in computing and situates modern AI as the culmination of years of progress. It posits that we have created a thing with processing power that outstrips human cognition and can capture aspects of reality beyond human detection. AI can now beat us, quite literally, at our own game, as seen in chess, where an AI trained on the rules of the game augments itself to make independent and more compelling moves beyond human comprehension. Similarly, AI can discover new antibiotics in record time by merely being exposed to fundamental principles. Chess and medical breakthroughs are frequent references in the book, demonstrating the extent to which AI would affect domains once reserved for humans.Chapter 2 is particularly delightful, with sentences brimming with such verve that one wishes it never ends. It explores centuries of sociocultural and sociopolitical forces that paved the way for AI and pranced through the evolution of human ingenuity, reason, and intellect. The chapter posits that AI's ability to upend every aspect of society surpasses the revolutions wrought by the printing press and electricity. These earlier technologies not only introduced new forms; they disrupted every aspect of society. The printing press bestowed new roles on the Western individual by wresting powers away from the Church and equipping the individual - facilitated by the Protestant agitations - with scholarly access to the divine. This psychological shift in the Western mind - sufficiently explored in Joseph Henrich's work, The Weirdest People in the World - launched the Renaissance, ushering in flourishing in arts, architecture, literature, and civic participation, ensuring the greatness of Michelangelo, Leonardo da Vinci, and others. The authors contend that AI is destined to follow a similar trajectory.The philosophical underpinnings of AI are captivating. The authors, probing AI's capture of reality, align it with Wittgenstein's view of making meaning through familiar connections rather than reducing reality to mechanistic explanations. Hence, the neural networks that inspire AI in mimicking the structure of the human brain place it far from mere computations of cause and effect, or garbage in, garbage out.“To enable machine learning, what mattered was the overlap between various representations of a thing, not its ideal — in philosophical terms, Wittgenstein, not Plato. The modern field of machine learning — of programs that learn through experience — was born.”Still, on philosophy, the book wonders if we are equipped to deal with our new fate. If AI can capture reality outside human conception, how do we retain our identity when perceptions would be determined by something beyond us? The authors concede that civilisation has been primarily created and sustained through the dynamics of Faith and Reason, and AI is designing a new form. It is a difficult notion to digest, since phenomena that thinkers and philosophers have grappled with, e.g., consciousness, divinity, nature/nurture, would become more challenging to comprehend.There is a lot of caution in the book. The authors warn that where nuclear weapon is the most dreadful of human arsenals, AI surpasses it by an order of magnitude. This apocalyptic view is further compounded by the difficulty of designing effective verification systems for a rather inscrutable technology. It is to wonder what we have gotten ourselves into. Human ingenuity has birthed a hybrid of saint and devil. Where nuclear weapon is under international regulations in which nations with nuclear capabilities are under the watchful eyes of post-WWII and post-Soviet accords, how do we police something so insanely hard to detect, easily distributed, and accessible? Nuclear deterrence has so far saved us from annihilation. How do we protect ourselves from something that possesses the capacity to "transform conventional, nuclear, and cyber weapons strategy"? This makes the book an entreaty, inviting governments, policy wonks, and military thinkers to convene and hash out red lines that would ensure responsible applications.For me, the positives outweigh the negatives if regulations are in place. And we must be careful to avoid stifling innovation under the guise of potential misuse. Moreover, as AI accelerates prosperity and instigates breakthroughs, how will it impact the global south? Will it leave a section of humanity behind while perpetuating historical patterns of economic inequities, a fact that Emad Mostaque of Stability AI has been vocal about? The book hints at it, but it would take a separate publication to articulate this concern.Overall, it is a delightful book written by those who should write about AI and society.

I quote these two passages from the book, which are spot on:1. “AI does not possess what we call common sense…” (p. 69)2. “Associations between aspects of inputs and outputs based on supervised or reinforcement learning are very different from human understanding… This brittleness is also a reflection of AI’s lack of self-awareness. An AI is not sentient. It does not know what it does not know.” (p. 69)To make these abstractions more concrete, I add this horrific example of Artificial Stupidity.Engineers at Boeing created software algorithms to help prevent the new 737 MAX from stalling during flight by automatically pointing the nose down. The software was needed to counteract the tendency of the MAX aircraft’s nose to pitch upwards during certain conditions, due to more powerful jet engines. Pilots at that time did not know about the MCAS software, nor about its dependency on a single external input. If that indicator sent false signals about the angle of attack, and the MCAS software did not know about the faulty data, it did what it was designed to do—point the plane’s nose downward. This was catastrophic in two cases, resulting in crashes that killed 346 human beings.In the end, AI is a software tool. All computer software operates on bits of information, and those bit collections can represent whatever we want them to. But we must always keep in mind that they are REPRESENTATIONS of reality, not reality itself. The signals from the faulty angle-of-attack sensor on the 737 MAX misrepresented reality. The pilots tried to correct the situation, but not knowing about the software’s design, they apparently did not disable MCAS or at least not soon enough, and it crashed those two planes.MCAS did what it was programmed to do. MCAS felt no remorse about this human tragedy. It did not even know.This is Artificial Stupidity. Discovery to Earth: Who will be the astronaut to disconnect HAL, the AI software run amok in Arthur C. Clarke’s novel (2001: A Space Odyssey?)Tools can be used for both good and bad purposes. I use a very good chef’s knife daily, and I keep its blade sharp and honed regularly. It does a great job in slicing and dicing vegetables. But it can cut flesh, when used carelessly. The tool is not good or bad in itself. How it is used by humans can be good or bad. Understanding how the chef knife is designed, and what it is capable of, is extremely important in order to use it properly and safely.Computers and the software they run (just bit patterns after all) are no different. Lack of understanding these fundamentals can lead us to Artificial Stupidity. Let’s hope it doesn’t destroy us as we learn what AI can do, especially when we don’t understand what AI is doing and why. This book does warn us about these issues, and for that reason it is worth reading. But it does not make this point strongly enough, in my view. I keep remembering The Sorcerer’s Apprentice, who used magic he did not understand (Walt Disney’s animation, based on Goethe’s original poem of 1797). That magic spell resulted in a situation run amok. In that fantasy, the wizard returned in time to stop the disaster from continuing.Who will be the wizards for disconnecting AI when it runs amok? Sadly, it’s already happening in our global social networks.