'Co-Intelligence'

For when the vibe is intellectual, curious, and neoteric.

‘Co-Intelligence’ by Ethan Mollick is one of the most compelling nonfiction books I’ve read in a long time – maybe ever. I finished it in a day and learned quite a lot about AI, and not just about how to optimize its use, but about how it “thinks” and works.

If you are like me and harbor a nascent fear of this thing that we’re creating based in watching “The Terminator” too many times growing up, it may comfort you to know that there are many things AI can’t do. It’s not self-aware. It can offer an approximate facsimile of human-like self-awareness when prompted, but it can’t offer genuine, specific insights into its programmatic decision-making processes. It can tell you what it “thinks” you want to hear about this or that based on the prompt you feed it, but not why it – the AI – reached a certain conclusion or said the things it said. It cannot, for example, write an accurately-sourced, academic, introspective book about itself and its relationship with humans – at least not yet. Thank goodness we have Ethan Mollick for that.

According to Mollick, the list of things AI can do is far more robust. It can generate “unique” ideas, if prompted correctly. It can draft and edit pieces of writing. It can offer critiques and analysis of…really anything and offer any perspective you can think of, from Genghis Khan to the Dali Lama. It can write effective code based on goals and parameters written in plain English. It can outstrip humans in academic tests, creative competitions and, sometimes, conversation.

The book itself offers a somewhat balanced perspective on AI, at once painting it as non-threatening, idiosyncratic, infinitely useful and also formidable, unnerving, and difficult to understand – a tool capable of out-pacing, out-witting, and out-competing humans. Never fear though: in this book Mollick offers tips and tricks for maximizing your own utility of and engaging effectively with AI, for making AI work for you…even if you are in one of those fields most likely to be threatened by its capabilities.

AI actually helped me with this post. I asked an AI for a photo of co-intelligence and it first gave me an unnerving sketch of two people in gas masks huddled together. I asked it to go back to the drawing board and it gave me the picture above. While still a little unsettling, it’s usable, and I had my hands on it within seconds…for free. Thanks AI!

According to Mollick, among the guiding principles of using AI is the rule that you must remain the human in the relationship. This means both that you cannot blindly rely on the answers AI provides as fact and that you control the output of AI by carefully crafting your input. AI is responsive to you.

If you’re worried that this book’s approach to AI is ‘too practical,’ Mollick does bookend all the practicum with discussions of the existential risks posed by AI. As a solution to some of the more grave risks, he is an advocate of pursuing now, when AI is in its earlier stages of development, for what is called “alignment.” “Alignment” can best be described as essentially ethics training for the machines. This would be not unlike the cadre in French parenting – a fixed framework through which the AI is allowed to interact with the world. That way, this will be less likely to kill us all when it becomes smarter than we are.

In the near term the bigger threat to the status quo of what makes a success in high earning fields is actually other humans who know how to use AI. Natural human skill, talent, and work ethic will mean very little with the great equalizer of AI available to everyone everywhere. Mediocre thinkers, writers and creatives will be able to hang with the best of them now, according to recent research.

But the takeaway is this: AI will be smarter than you one day, and it’s probably already better than you at your job (depending, of course, on what your job is. Football players are safe…but their coaches might not be). Here, Mollick offers a practical approach for the everyman to start interacting with AI, to get in on the ground level as it were, so that they can know best how to use it, how it’s changing, and how to keep up. It doesn’t take the kind of brooding, doomsday approach or the idyllic, reverent perspective that is expected in the public discourse surrounding AI. Rather it focuses on a well-researched, utilitarian view into AI practicum, which can serve as a helpful guide for any AI novices out there looking to dip their toes into this large, mysterious, and possibly threatening ocean.

Buy the Book

$17.79 at Amazon