God AI: Boom or Doom? What to Expect When Machines Outsmart Us, Toby Walsh (La Trobe University Press, an imprint of Black Inc., 2026).
For me, AI began with books by legendary AI thinkers Nick Bostrom and Stuart Russell, read in the ‘park years’ during lunchtime at Flagstaff Gardens in Melbourne, years before I encountered the tools. The language of Armageddon arrived before the interface! This is a strange introduction: learning to fear AI before discovering whether it could answer a question usefully. Toby Walsh’s God AI revisits this “AI Safety” neighbourhood, where humanity is approaching a threshold, and civilisation apparently needs another computer scientist to explain the future!
Bostrom helped formalise the modern existential-risk framework in 2002 and popularised superintelligence through his 2014 book; he did not invent every anxiety about thinking machines, but the AI industry likes to parrot him. AI safety addresses genuine problems of AI control and reliability. Still, its extinction-focused literature can feel less like an open inquiry than an established pile: familiar fears, increasingly extravagant stakes from an old but narrow field. The whole species is invited, although the conversation often sounds remarkably parochial. The claim to speak for all of us is what needs interrogation: whose history, whose institutions, whose assumptions about progress, and whose authority to decide which future counts as desirable. Is AI really that important?
The vocabulary deserves a rough audit. From my searchers, across the chapters, I estimate ‘humanity’ appears 89 times, ‘existential’ 47, ‘extinction’ 50 and ‘singularity’ 26; ‘existential risk’ and its plural appear 23 times together, ‘civilisation’ and its plural 15, with another two uses of ‘civilisational’.
Of course, frequency alone proves very little, but the effect is faintly ridiculous. Every discussion seems to require a species, an epoch and somewhere suitable to announce the end of history. Occasionally, a smaller noun would do! Walsh is an engaging, knowledgeable guide, and at times, an effective puncture of AI inflation. He is critical of the AI world while remaining thoroughly of it. The sceptic and the enthusiast keep exchanging the microphone, sometimes before either has finished the sentence!
The historian down the corridor might be able to help. ‘Civilisation’ is not a benchmark with a leaderboard, nor is the human past a training ground for Silicon Valley. Walsh does draw on writers and philosophers, including Dreyfus and Hobbes, so this is not a book without humanistic reference points. The problem is that borrowing their names does not necessarily justify the technological framework doing the borrowing.
Walsh is most persuasive when critiquing LLMs. He argues that much of their apparent intelligence reflects sophisticated information processing rather than understanding. This does not make them merely search engines, but it does raise an uncomfortable educational question: why do universities (especially in the sciences) so readily treat the reproduction of information as evidence of intelligence? This scepticism also informs his uncertainty about whether current language models can reach artificial general intelligence; the book spends considerable time entertaining the Singularity, recursive self-improvement and extinction scenarios, but in this reader’s mind, their presentation gives them more substance than they warrant.
Medicine provides a more grounded prospect: Walsh anticipates substantial benefits where computational analysis can help explain biological complexity. His discussion of human labour is less coherent, reduced labour promises liberation, yet the loss of meaningful employment also threatens human purpose. It requires attention to how work, income and social recognition are organised. Greater technical capability does not, by itself, settle those questions.
The conclusion reproduces this tension between scepticism and speculation. Walsh identifies the religious character of AGI discourse, then examines AI through divine attributes such as omniscience and immortality. The sermon is criticised, but the organ keeps playing. God AI is worth reading against its own rhetorical current: its strongest passages concern institutions, power and responsibility. Machines can remain machines, and understanding their consequences requires better social analysis, not just more pile-on.
References
Bostrom, N. (2002). Existential risks: Analysing human extinction scenarios and related hazards. Journal of Evolution and Technology, 9(1). https://nickbostrom.com/papers/existential-risks/nickbostrom
Bostrom, N. (2014). Superintelligence: Paths, dangers, strategies. Oxford University Press.
Walsh, T. (2026). God AI: Boom or doom? What to expect when machines outsmart us. La Trobe University Press.
AI declaration
This review was developed with assistance from Perplexity AI, which generated a draft from my close reading notes and later instructions. The personal reading experiences and critical perspective remain from my notes.
Leave a Reply