Book Review: Superagency by Reid Hoffman + Greg Beato

This week's review is on Superagency: What Could Possibly Go Right with Our AI Future by Reid Hoffman and Greg Beato.

Hoffman is the cofounder of LinkedIn and has investments in a number of AI-related firms, while also serving on the boards of Microsoft and other tech giants. His position is key to understanding the truly insane takes present in this book. I use that phrase on purpose - the book presents a worldview that seems ungrounded in our current reality and in a manner drawn to create controversy or ridicule doubters.

Bad cover.

Superagency begins with an introduction centered on the idea of "new technologies...spark[ing] visions of impending dehumanization and societal collapse" (xi) before examining this in the context of the printing press, telephone, car, automation, and AI. Right away, the text discards the real negative effects of these inventions - the printing press in fact destabilized society, leading to massive religious wars in the immediate aftermath of the Gutenberg press and centuries of misinformation, concerns over truth, and the best way to share information (concerns Hoffman clearly has given his political advocacy). The telephone actually did change societal relationships - just see the impact of the cellphone. Cars have fueled suburbanization, societal atomization, and individualism in startling and unseen ways since their widespread adoption. Automation, of course, still has lingering effects on our economy, politics, globalization, and trade. None of these concerns are taken seriously in the first pages of the text, nor as they are referred to throughout the book. This dismissal permeates every concern of AI the book pretends to address.

The text proceeds through a vision of AI's impact in various ways, tossing in historical allusions akin to those on the first page, likewise devoid of true historical understanding or societal implications. AI doubters are "gloomers" or "doomers," akin to wanting to be the Donner Party traveling to the West Coast instead of taking a car* (yes, this is a real analogy deployed, see pages); or anti-progress Luddites who desire economic ruin as the rest of the world pursues an AI-powered future.

Superagency is supposed to be focused on "can we continue to maintain control of our lives, and successfully plot our own destinies?" (11) - a question that the authors feel is stupid to even ask. AI will "increase your agency, because it's helping you take actions designed to lead to outcomes you desire" (13). There is little discussion of the widespread stolen data for training nor the creative slop AI has produced that overwhelms Spotify, YouTube, and more.

In the second chapter, Hoffman and Beato cast the AI constituencies into Doomers (ASI takes over), Gloomers (AI isn't good but in realistic ways), Zoomers (AI needs no limits and its impacts will be awesome), and Bloomers (AI should be developed iteratively). I believe Hoffman and Beato cast themselves in the last camp, but they live in third. These camps are referenced at times but don't form a coherent framework within the episodic and hand-wavy nature of this book.

They make the case for the necessity of trust, while extolling possibilities in counseling (Chapter 3), the promise of Networked Data (4), Benchmark considerations (5), the necessity for market-driven regulations rather than government-driven ones (6) [ignoring seat belts and other car safety mandates], support for people just like GPS offers (7), a discussion of the way law and norms interact with AI possibilities (8), networked AI's promise (9) such as in cases with Covid (more below), and visions for future AI use (10-11). The book purports to promote the idea of "human agency... shared data and knowledge promoting individuals... and that innovation and safety go hand-in-hand" (page 229).

It's not that these chapters aren't at times compelling - it's just that they seem set in an alternate reality. Let's take two examples - drawn from Chapter 4 where Superagency seeks to engage with critiques of AI and capitalism and Chapter 9 where the authors present AI's promise as a positive.

They acknowledge the idea of surveillance capitalism (Chapter 4) and the flaws in Zuboff's account - while neglecting the genuine critiques of Google and its monopoly (and how that has funneled AI growth and development). They present an image of this data as not "usurping value from users...[but one where] a mutualistic ecosystem of developers, users, and content creators whose interactions and contributions collectively enrich the lives of billions of people every day" (78). This creates surplus value - one that bases itself on how users feel the value (in terms of what they would need to be paid to give it up) - a good reminder that as Ted Gioia says only tech companies and drug addicts call their customers "users."

In Chapter 9, Covid-19 is presented as a perfect example of why a country might adopt mass AI-driven surveillance where "such a system will maximize its citizens' freedom from illness, lockdowns, and economic disruption and preserve their freedom to work, travel, and gather without fear" (198). Of course, we now know that the extreme lockdowns, especially of children, did not offer protection worth its damages - and that efforts to ignore such guidelines and frequent controversial places such as parks and nature trails - were a net positive. The authors likewise do not engage in the possibility of the government using such a system for non-pandemic purposes like tracking Autistic children or controversial political beliefs.

AI Generated, of course.

Lack of serious engagement or efforts to see any other side of the issue is what makes Superagency such an ill-considered project. Its hype for AI seems unrealistic, untethered, and, ultimately, unhelpful.

Superagency: What Could Possibly Go Right with Our AI Future by Reid Hoffman and Greg Beato.

Rating: 2/5 Stars

Good For: Those wanting bad comparisons and rose-colored, AI-generated glasses.

Best nugget: History struggles to adopt things - sometimes that can hold us back from important things. Sometimes it doesn't. Sometimes people disagree on the interpretation of those things - especially if they are financially involved.

*"So, who, really, is freer? We citizens of the twenty-first century with our seat belts, speed limits, and comprehensively surveilled roads, who can make that journey in a day and a half? Or those rugged individualists in the Donner Party who had to resort to the ultimate form of communism to survive a wrong turn they made on the untrammeled frontier?" (190)

Please note: As an Amazon Associate I earn from qualifying purchases. However, I am not paid to provide reviews or use content.