Taming Silicon Valley

July 22, 2024
“We... need checks and balances if we are to get to a positive AI future,” according to this discerning treatise. Marcus (The Algebraic Mind), founder of the machine learning company Geometric Intelligence, posits that AI will never be able to “save society,” as tech companies claim, given that it’s designed to predict word sequences and is incapable of understanding the text it generates. To illustrate AI’s incapacity to reason, Marcus notes that ChatGPT told him “1kg of bricks and 2kg of feathers weigh the same.” The technology is also dangerous, Marcus contends, showing how he circumvented restrictions on asking the chatbot for bomb-building instructions by changing the formatting of his request. His proposed regulations for ensuring the responsible development of AI include the usual calls for artists to receive compensation for work fed into language models and for greater privacy protections (Marcus posits that every user should have “clarity about how their data are used,” a choice to opt out of data collection, and “a cut of the profits from using their data”). However, Marcus also makes a persuasive case for more ambitious policies, arguing that because AI will likely render some jobs obsolete, the federal government should consider instituting a universal basic income to cushion those affected. Shrewd and levelheaded, this provides plenty to ponder.

Starred review from August 1, 2024
An esteemed academic shows why AI needs firm regulation, which can only come through public pressure. With the release of ChatGPT, AI burst into the public consciousness, accompanied by voices sounding warnings about its dangers. Marcus, NYU professor emeritus of psychology and neural science and the author ofThe Algebraic Mind andKluge, delivers a significant entry in the rapidly expanding literature on the positives and negatives of AI. "What we have now is a mess, seductive but unreliable," writes the author. "And too few people are willing to admit that dirty truth." In the early chapters, Marcus recounts the mistakes of AI creators and systems. Some of them are comical, and many of them are frightening, but the errors should be avoidable. The author argues that current AI systems were released prematurely in a race for market share. Moreover, the companies that promote them are now more interested in making money than fixing the problems. In other words, things will get worse before they get better. Big Tech has been effective at stopping attempts at regulation of AI, through political alliances, clever marketing, and vague promises of self-regulation. The author argues convincingly for the necessity of public pressure to spur regulators into action, with a campaign that would lead to laws based on transparency and accountability. It will be a long and difficult process. "I suggest we start with one simple act," writes Marcus. "Let's stand up for all the artists, musicians, and writers we love, and boycott Generative AI companies that use their work without compensation or consent." This seems like a good place to begin; some of the author's other proposals sound more hopeful than realistic. Nevertheless, this book is an important contribution to a crucial debate. With passion and expertise, Marcus sounds a strong call to action.

COPYRIGHT(2024) Kirkus Reviews, ALL RIGHTS RESERVED.