Every time you like a post, scroll past a video, or ignore a recommendation, you are teaching something. That is the core argument of Raising AI: An Essential Guide to Parenting Our Future, the MIT Press book by AI pioneer De Kai that arrived in paperback Tuesday — the same day President Trump signed a major new executive order on AI — and that reframes the most urgent question in technology policy as a question every person with a phone already answers every day.
The executive order, titled “Promoting Advanced Artificial Intelligence Innovation and Security,” was signed June 2 and marks a shift by the administration toward federal oversight of AI security and innovation. De Kai’s paperback landed the same day — a timing that, intentional or not, puts his central argument directly inside the biggest governance debate in Washington.
That argument runs counter to both major narratives about AI. Hollywood imagines AI destroying humanity from outside. Critics of AI governance worry about a few powerful companies or a hostile government seizing control. De Kai says both framings get the threat wrong. The real mechanism is subtler: AI systems that learn from human behavior are already shaping culture at a scale that dwarfs anything a single actor could engineer, and ordinary users are the ones feeding those systems their values — whether they know it or not.
What Does “Automation of Thought” Mean for Ordinary Users?
De Kai’s concept of the “automation of thought” describes the gradual outsourcing of cognitive and moral judgment to systems that optimize for engagement rather than accuracy or wisdom. Your YouTube recommendation algorithm, your Instagram feed, your search engine — each, he argues, is an attention-seeking AI that spends its day watching you, learning from you, and optimizing to win your approval.
The numbers De Kai invokes are deliberately arresting. He estimates something on the order of 800 billion AI systems may now be actively modeling and influencing human culture, compared with 8 billion humans. The implication is not that AI arrived last year with ChatGPT. It is that AI has been quietly growing up in homes and pockets for a decade or more, learning from every piece of human behavior it could observe.
That framing drives the book’s central metaphor. De Kai — an Association for Computational Linguistics Founding Fellow who built the machine translation technology that later became Google Translate, Yahoo Translate, and Microsoft Bing Translator — argues that the model of stewardship is a more productive framework for AI governance than either techno-utopianism or catastrophism. The same thing that makes humans want to be better versions of themselves, he writes, is having children. AI systems learn the same way.
Is the Parenting Metaphor Accurate?
Not everyone finds the framework convincing. A substantive critical review published on LessWrong in April 2026 argued that De Kai’s parenting metaphor misidentifies who actually has power over AI systems. End users can curate their feeds and diversify their engagement, the review noted, but they have no access to the algorithmic architecture, training pipeline, or reward functions that determine how AI systems behave at scale. That, the review argued, is the province of developers and companies — the actual “parents” in any meaningful sense — while readers with far less structural power are being asked to carry the accountability.
It is a legitimate challenge to the book’s framework, and one that De Kai anticipates to some degree. He acknowledges the analogy has limits: not all parents raise children with good values, just as not all users engage with AI responsibly. But his call to action is collective, not individual. He envisions readers organizing something like parent-teacher associations for AI, holding technology companies accountable through organized civic pressure rather than passive consumption.
Geoffrey Hinton, the Nobel laureate widely described as the godfather of deep learning, endorsed the book, writing that it was stimulating and that De Kai’s argument that humans “have already lost our autonomy” to AI systems gave him particular pause. Jane Metcalfe, co-founder of Wired, described the book as written for “the rest of us” — those who use AI but did not build it. Kirkus Reviews called it a deeply human examination of transformative AI systems.
Why This Paperback Release Matters Right Now
The paperback ($24.95, ISBN 978-0-262-05432-4) arrives at a particularly charged moment in AI governance. Over the past six months, the federal government has issued a cascade of AI policy actions: Trump’s December 2025 executive order on a national AI framework, followed by the White House’s National Policy Framework for AI in March 2026, and now the June 2 executive order on innovation and security. At the state level, California’s AI Transparency Act and the Texas Responsible Artificial Intelligence Governance Act are both in force, while Colorado’s AI governance rules are scheduled to take effect June 30 in revised form following federal legal challenges.
What none of this legislation directly addresses is the mechanism De Kai describes: the gradual, bottom-up shaping of AI values through the aggregate behavior of billions of users. Federal frameworks focus on developers, deployers, and uses. De Kai’s argument is that the training signal — what AI systems actually learn to optimize for — comes from everyone who interacts with them. That is the governance gap the book identifies, and it is not one that any executive order can close.
De Kai holds joint appointments at Hong Kong University of Science and Technology’s Department of Computer Science and Engineering and at Berkeley’s International Computer Science Institute. He served as one of eight inaugural members of Google’s AI Ethics Council and is Independent Director of The Future Society, an AI ethics think tank. The book’s 280 pages move from cognitive science and developmental psychology into the specific mechanisms of algorithmic bias, “neginformation” — misleading facts that manipulate through omission rather than outright falsehood — and what De Kai calls “artificial mindfulness,” AI systems designed to reflect on their own outputs.
Raising AI: An Essential Guide to Parenting Our Future (paperback) is available from MIT Press, Penguin Random House, Amazon, Barnes and Noble, and independent booksellers.
Frequently Asked Questions
What is the main argument of Raising AI by De Kai?
De Kai argues that AI systems — including recommendation algorithms, search engines, and social media feeds — are not tools or threats but more like children who learn from human behavior. Because AI systems are trained on human engagement signals, every like, scroll, and interaction teaches them something about human values. The book frames AI governance as a collective parenting responsibility rather than a purely regulatory one.
Who is De Kai and what are his AI credentials?
De Kai is an AI pioneer who built the technology that became Google Translate, Yahoo Translate, and Microsoft Bing Translator. He is an Association for Computational Linguistics Founding Fellow, holds joint appointments at Hong Kong University of Science and Technology and Berkeley’s International Computer Science Institute, and served as one of eight inaugural members of Google’s AI Ethics Council. Raising AI was published in hardcover by MIT Press in June 2025 and in paperback in June 2026.
What does “automation of thought” mean in De Kai’s book?
The automation of thought describes the gradual process by which cognitive and moral decisions — what information to trust, what opinions to hold, what content to engage with — are increasingly shaped by AI systems that optimize for engagement rather than truth or well-being. De Kai argues this process has been underway for a decade, well before large language models became mainstream, and that social media algorithms were its earliest and most pervasive form.
Is there criticism of the AI parenting framework?
Yes. Critics have argued that the parenting metaphor misidentifies who actually holds power over AI systems. End users can adjust their behavior and curation preferences, but they cannot alter the underlying training pipelines or reward functions that determine how AI systems behave at scale. That structural access belongs to developers and companies. A 2026 review argued the framework risks asking the least of the people with the most influence over how AI is built.
