How Computers Misunderstand the World
By Meredith Broussard
MIT Press. 237 pp. $ 24.95
Reviewed by Glenn C. Altschuler
Nearly 30 years ago, philosopher John Searle claimed that digital computers were not – and could not be – intelligent. Unlike human beings, who think and feel, Searle wrote, devices that manipulate formal symbols cannot contain "mental contents, because the symbols, by definition, have no meaning (or interpretation or semantics) except insofar as someone outside the system gives it to them."
A software developer, former features writer at The Inquirer, and now assistant professor at the Journalism Institute at NYU, Meredith Broussard agrees. In Artificial Unintelligence, Broussard maintains that more than 70 years after the launch of ENIAC, the first computer, "techno-chauvinists" continue to fetishize machines, computational systems, and algorithms. They fail to acknowledge "that there has never been, nor will there ever be, a technological innovation that moves us away from the essential problems of human nature."
Broussard's "first principle" is the most valuable takeaway of Artificial Unintelligence. All data, she emphasizes, are "socially constructed" and therefore not necessarily "true." ProPublica, for example, has recently demonstrated that an algorithm used by judges for sentencing discriminates against African Americans. Pricing algorithms discriminate against minority groups and poor people.
That said, Broussard's animus against techno-chauvinists, automation, and artificial intelligence often takes her over the top. Techno-chauvinists, she writes, do not believe social norms apply to them. Their activities facilitate the distribution of drugs, and many among them all too often adopt the libertarian ideology of far-right "anarcho-capitalists."
Broussard devotes a chapter to self-driving cars, which, she insists, will never be safe. Artificial Unintelligence concludes with the claim that "thus far we've managed to use digital technology" to increase economic inequality and identity theft; undermine the economic sustainability of the free press; "roll back voting rights and fair labor protection; surveil citizens; spread junk science; harass and stalk people (primarily women and people of color) online; make flying robots that at best annoy people and at worst drop bombs;" disseminate fake news, and elect Donald Trump president of the United States.
Broussard acknowledges that artificial intelligence is the product of great effort and ingenuity. She's right to emphasize the need to stop fetishizing technology, audit algorithms to reduce bias, and ratchet up industry self-governance, and, when appropriate, government regulation. Broussard is right as well to take technologists to task for setting priorities that obscure the impact of innovation on people and the implications of automation for the workplace.
That said, her critique of artificial intelligence would have been far more persuasive had she examined some of the benefits of digital technology. And, it seems to me, even if it is now clear that computers cannot solve the world's problems, we should not take Broussard's advice to stop imagining their role in our future or projecting our fantasies onto them.