Don’t write a book on artificial intelligence. You’ll regret it.
A.I. is hotter than hot. If you have a unique point of view, by all means write an article about it.
But I don’t care how much you know, this is no time to write a book about it.
Five kinds of A.I. books and why you shouldn’t write them
There are more than 20,000 titles available on Amazon with “artificial intelligence” in the title or subtitle. That creates a serious problem for you as you attempt to differentiate your book from all the rest. But that’s not your only problem. I’ve never seen a technology growing and changing more rapidly than AI is right now. That means whatever you’re thinking of writing will rapidly become obsolete and irrelevant. And, most likely, very soon, obviously wrong as well.
I’ve written about how to write about rapidly evolving technologies before. All those strategies remain relevant. It’s just that for AI, they aren’t powerful enough to keep up with the pace of change.
Here are five kinds of books you shouldn’t write:
- The rapidly published how-to book. You can get a self-published how-to book out in three or four months. But if it mentions anything about how to use AI tools like ChatGPT or Meta LLaMa or Perplexity, it’s going to become obsolete extremely rapidly (maybe before it’s even published). It’s not just a question of your how-to instructions becoming out of date. The products and use cases you describe may become irrelevant, too. Imagine a book of advice on COVID-19 that was published in June of 2020 — everything it says is now obsolete. The same thing is going to happen to what you write, only much faster.
- The textbook. Whatever you want to write about best practices for, say, coding with AI, writing with AI, managing organizations with AI, statistical analysis with AI, and so on is going to be wrong soon. A textbook might take two years to publish. Do you really think your instructions will have the slightest relevance to the students of 2027? They’ll be laughing at you and any professor out-of-touch enough to recommend your book.
- The book of predictions. You think you know where A.I. is going? Ha! Among the unknowns are how managers will treat it, how governments will regulate it, how big tech companies will transform it, how it will affect every software product ever conceived or implemented, what backlash there might be from consumes, whether its use of training data will be outlawed, where the energy to power it will come from, and, frankly, whether and how it will achieve sentience and tell us what to do. All of these shifts are moving in unpredictable directions and interacting with each other. This is a chaotic system, and chaotic systems are impossible to predict — it’s like predicting the weather on today’s date a year from now. To be clear, books of predictions always become obsolete because change happens slower, faster, and in different directions that analysts predict — but even so, books of predictions about trends that move at normal rates of speed often remain relevant for five years or so. In this case, your predictions will likely diverge from reality in six months. Given a typical 18-month publication cycle for a traditionally published hardback book, your predictions will be laughable right around the time they appear.
- The management book. Ah, the ever-popular “this trend is coming, so here’s how managers and leaders should prepare for it” book. But for a book like that be useful, it assumes first, that you can define the things to manage, and second, that they are manageable at all. When it comes to AI, both of those assumptions are questionable.
- The long big-think strategic view. One way to get past the problem of rapidly developing technologies is to take a strategic view. In a book like that, you don’t write about specific technologies, you write about trends and how to position yourself for them; those strategic statements have a longer shelf-life than any specific technology product. I’ve cowritten two of these books. My book on business strategy for social media, written at the dawn of social media and revised three years later, remained relevant for about five years. (The MySpace chapter is how a historical artifact.) My book on business strategy for mobile, written at the dawn of the smartphone age, remained relevant for about three years. A similar book on AI will remain relevant for a year if you’re lucky. Given the level of effort required to write and publish such a book, that’s a pretty short shelf-life.
I know many of you will ignore my advice. You’re the world’s foremost expert. You have a unique angle. You work for OpenAI. You’re Satya Nadella or Andrew Yang. You invented large language models.
I salute you. Unfortunately, you’re hosed.