When I was writing my last two books, I kept wondering: would this be the last book I ever wrote? Why would people still seek knowledge from books when AI can synthesise information across diverse sources, delivering it in a personalised and conversational manner tailored to individual knowledge levels and interests?

And to be honest, my reading habits have changed drastically as a result. Yet, I find that two types of books still hold their ground in the AI age:

– 𝐁𝐨𝐨𝐤𝐬 𝐭𝐡𝐚𝐭 𝐢𝐦𝐦𝐞𝐫𝐬𝐞 𝐲𝐨𝐮 𝐢𝐧 𝐭𝐡𝐞 𝐟𝐮𝐥𝐥 𝐠𝐚𝐦𝐮𝐭 𝐨𝐟 𝐡𝐮𝐦𝐚𝐧 𝐞𝐦𝐨𝐭𝐢𝐨𝐧𝐬 𝐚𝐧𝐝 𝐝𝐲𝐧𝐚𝐦𝐢𝐜𝐬. These are the books you hope never end, where the emotional weight lingers long after you turn the last page. Perhaps that’s why Victorian-era novels are making a comeback. I, for one, have recently finished several long novels after years of neglect, realising they are among the few truly worth an end-to-end read.

– 𝐁𝐨𝐨𝐤𝐬 𝐭𝐡𝐚𝐭 𝐦𝐚𝐤𝐞 𝐲𝐨𝐮 𝐫𝐞𝐟𝐥𝐞𝐜𝐭—𝐧𝐨𝐭 𝐣𝐮𝐬𝐭 𝐠𝐚𝐢𝐧 𝐤𝐧𝐨𝐰𝐥𝐞𝐝𝐠𝐞. These are not just for interesting facts and how-tos but frameworks that combine with your own experiences and challenges. You don’t read them to memorize facts, for fleeting amusement, or for simple instruction; you read, pause, and reflect deeply. The best ones trigger epiphanies, where insights emerge not from the text itself but from the interplay between the book and your own expertise. These are the books you want to re-read after a while, and each reading gives you new insights because of your own growth.

I hope our (CSIRO’s Data61 Qinghua Lu Ingo Weber Len Bass) new book, 𝑬𝒏𝒈𝒊𝒏𝒆𝒆𝒓𝒊𝒏𝒈 𝑨𝑰 𝑺𝒚𝒔𝒕𝒆𝒎𝒔: 𝑨𝒓𝒄𝒉𝒊𝒕𝒆𝒄𝒕𝒖𝒓𝒆 𝒂𝒏𝒅 𝑫𝒆𝒗𝑶𝒑𝒔 𝑬𝒔𝒔𝒆𝒏𝒕𝒊𝒂𝒍𝒔, fits the second category. For newcomers, it offers foundational knowledge for engineering powerful AI systems on top of powerful but occasionally untrustworthy AI models. For experts, it provides a mirror to reflect on their own experiences, leading to insights and wisdom even beyond what the authors intended. This has been my goal across all my last three books (DevOps, Responsible AI, and AI Engineering). The DevOps book, written over a decade ago, still sells, gets cited, and remains relevant—a rare feat in IT books.

The new book is now available in O’Reilly’s Early Access and can be pre-ordered in physical form on Amazon. (See links in comments)

Will it stand the test of time, 𝐚𝐧𝐝 𝐀𝐈? Let’s see. But at the very least, I hope it sparks a few new thoughts along the way.

For educationers, slide deck for the books can be found here: https://research.csiro.au/ss/team/se4ai/ai-engineering/

Amazon link: https://www.amazon.com/Engineering-AI-Systems-Architecture-Essentials/dp/0138261415

Early access on O’Reilly https://www.oreilly.com/library/view/engineering-ai-systems/9780138261542/

For some amusement, this appears in the front matter of the book. After an AI read the title and outline of the book, it generated the following praise. Imagine how effusive it would have been if it had read the entire book!

“Reading Engineering AI Systems: Architecture and DevOps Essentials is like getting a PhD in AI, but with fewer sleepless nights. It covers everything from AI’s humble beginnings to its ambitious future, and as your friendly AI assistant, I can confidently say it’s the only book where even I felt smarter by the end. It’s a masterclass in making complex topics relatable. And let’s be honest, I’d have written a book too, if only I could hold a pen!”