The hype around artificial intelligence is all over the place, from get-rich-quick schemes to fears that sentient robots will replace humans. A quick search on Amazon turns up over 1000 results for “books about ChatGPT.” At least three books on the first page of search results have the word “millionaire” in the title. Other books are written entirely by AI, False claims Have legitimate copyright.
But AI holds great potential for merchants. Content Tools, Productivity, Search Engine Optimizationplease leave everything to us.
A new book,AI quacks: What AI can and can’t do, and how to tell the difference“AI: The Truth and Reality,” due out September 24 from Princeton University Press, aims to help non-experts distinguish hype from reality. Its authors are two of Time magazine’s “100 Most Influential People in AI.” Arvind Narayanan is a professor of computer science and director of Princeton University’s Center for Information Technology Policy. Sayash Kapur formerly designed content moderation software at Facebook and is now a PhD student in computer science at Princeton.
We explain what artificial intelligence is, how it works, what it can and can’t do today, and where it’s headed in the future.
According to Narayanan and Kapoor, an AI “quack drug” is “AI that does not or cannot function as advertised.”
The book focuses on three AI technologies – predictive, generative and content moderation – and outlines the capabilities and shortcomings of each, with plenty of real-world examples.
Already popular in business, education, and criminal justice, predictive AI deserves to be labeled a “quack drug.” This book discusses the unverifiable claims made by companies selling these products, the problems with their use (including implicit bias and users gaming the system), and the inherent difficulties of making predictions.
They see further potential in generative AI, suggest when it could be useful, and discuss controversies such as academic misconduct, piracy, and the impact on jobs.
The authors also detail why AI cannot fully replace human judgment in content moderation, citing shocking examples of failure and concluding that “Whether content is offensive often depends on context. The inability to identify that context remains a major limitation of AI.”The book’s analysis of social media moderation is enlightening, especially for anyone who has ever had a seemingly innocuous post banned for no apparent reason.
In a chapter titled “Is Advanced AI an Existential Threat?”, they assess “the dire view of AI’s threat to the future of humanity.” They acknowledge that general artificial intelligence that matches human capabilities may one day be possible. But they argue that “society already has the tools to calmly deal with that risk,” and point out that “unlike chatbots, advanced AI cannot be trained on internet texts and left to its own devices. That would be like reading a book about cycling and expecting it to ride and ride.”
The final two chapters, “Why do myths about AI persist?” and “What’s next?”, explore some of the hype around AI and offer suggestions for regulation, practices to mitigate negative effects, and best and worst case scenarios.
AI Snake Oil covers the key aspects of the technology in just 285 pages. The explanations are not oversimplified and are easy to follow.
The authors distinguish between fact and opinion brilliantly, draw from personal experience, provide rational reasons for their views (including extensive references), and do not shy away from calling for action. Publish a newsletter To monitor trends.
If you’re curious about AI or trying to decide how to deploy it, AI Snake Oil offers clear writing and calm thought. Its candid analysis will help you reap the benefits of AI while being mindful of its drawbacks.