Chapter Summaries
The Social Impact of AI. This chapter discusses the positive impact of AI on our everyday lives. It explains that every one of today’s AI systems are one-trick ponies that have no human-level intelligence. Last, it sets up the rest of the book to explain how these systems work, why they will not evolve into human-level intelligence, and how AI will impact the future of humanity.
Fears Worth Having. Though AI will not evolve into the evil robots or killer computers of science fiction lore, AI can still have a negative impact on society – if we let it. This chapter discusses the legislative and political issues that are arising from the use of AI to create semi-autonomous capabilities that can be used in weapons of war, autonomous vehicles, and many other applications.
Brief History of AI. Over the last 60 years, there have been three periods of high optimism around AI including the current one. In between, there have been two “AI winters”. This chapter takes the reader through the history of the three AI hype cycles and two AI winters and explains what is different about the current incarnation of AI.
Employment. AI will not take all our jobs because we will not see human-level AI in our lifetimes. However, AI will have an impact on employment. This chapter reviews the impact of technology advancements in general on employment, asks the question of whether the impact of AI will follow the historical patterns, and summarizes the thoughts of experts on the topic.
Supervised Learning. This chapter starts by explaining simple supervised learning systems and working up to credit card fraud detection and the Cambridge Analytica scandal. In each case, the result of supervised learning is a function that can do one and only one thing and has no human-level intelligence.
Deception. One of the downsides of AI technology is that it provides a powerful tool kit for deception. Bad actors can use these tools to shake the very foundations of a free society through the manipulation of elections. Widespread misinformation, digital impersonation, and uncanny human-like robots are now (somewhat) possible. This chapter discusses legislative changes necessary to reign in these bad actors.
Unsupervised Learning. This chapter explains unsupervised learning, describes how deepfakes work, and explains how AI can be used to generate fake news. The bad news is that today’s AI can create scarily good deepfakes. The good news is that it can only create readily detectable fake news and this is likely to be a long-term limitation. This chapter also explains why unsupervised learning systems can do one and only one thing and have no human-level intelligence.
What Drives Self-Driving Cars. Self-driving cars use many of the AI technologies explained in this book. This chapter explains the history and technology behind autonomous vehicles. It explains how self-driving cars work and explains why it will be difficult if not impossible to build safe self-driving cars.
Reinforcement Learning. This chapter explain how reinforcement learning works and how it was used to beat the best Go players in the world. The chapter also explains how researchers are using reinforcement learning to build robots. Here also, the resulting systems are functions that can do one and only one thing and have no human-level intelligence.
Privacy. George Orwell’s dystopian novel Nineteen Eighty-Four describes a future society in which the government continuously monitored everyone’s actions and conversations. Today’s AI technology made that level of monitoring possible, and society needs to cope with the consequences.
Deep Learning. This chapter explains deep supervised learning and describes how machine translation, speech recognition, and facial recognition work. In each case, the result is again a function that can do one and only one thing and has no human-level intelligence.
Natural Language Processing. This chapter opens up the hood on IBM’s Watson computer system beat two Jeopardy! champions in 2012, explains how chatbots like Siri and Alexa work, and explains how Microsoft’s system that “reads better than humans” works. All these systems are created with clever programming tricks and have no human-level intelligence.
Thinking and Reasoning. This chapter examines AI research on building computers and robots that can think and reason. The chapter explains why there has been no real progress in this endeavor.
Discrimination. This chapter discusses how AI contributes to discrimination against minorities and invasion of privacy and discusses possible approaches to regulation, many of which are already in process. The chapter also explains why, even though the popular press blames AI for these issues, these issues would exist if AI had never been invented.
Artificial General Intelligence. This chapter summarizes the current state of AI and concludes that, while AI has produced a great deal of technological wizardry, every AI system is a one-trick pony with no more real intelligence than a toaster. Moreover, none of today’s AI technologies will lead to human-level intelligence. While it’s not possible to state definitively that human-level intelligence will never be achieved, we have no more idea of how to build it today than we did in 1980. That means it’s about as likely to be achieved as time travel. Regarding the development of human-level intelligence, the chapter concludes that “…we’re at the starting gate. Again.”
AI will not take over the world – unless we let it. The epilogue summarizes the author’s thoughts on how we should approach the negative social issues associated with AI. First, we need to stop worrying about Terminator-like scenarios. These aren’t happening and ascribing thinking and reasoning capabilities to computers and robots unnecessarily complicates the task of finding solutions to social issues. Second, we need to stop thinking about AI as a unitary cause of social issues. Each of the social issues discussed in this book is a separate issue and requires a distinct solution.