Human
Workers of the World - Unite!
After the constant hype around super-intelligence, it makes a refreshing change to turn to a book that considers AI in the way that we should consider most technology: something that can make our (working) lives considerably better, if we learn to use it properly. In most cases, this will require some changes in our habits: Mollick seems to be the kind of optimistic human being who thinks that people are generally inventive and flexible enough to learn to change (and, in general, I agree with him), although part of me still worries that part of the outcome of an inevitable push towards AI will be automation and job losses, with considerable anguish for people around the world.
Mollick is not immune to this, and there is considerable humanity at the heart of this book. If I have one complaint - and it is, unfortunately, a familiar one - it is that I wish an editor had been a bit more ruthless on cutting this down to size. Parts of this feel like padding, although some of that is doubtless due to the fact that I've spent considerable time studying what AI is and how it works, while a manger looking for ways to transform their team may very well appreciate the always lucid primers. Nonetheless, there are parts that would have been better as a paper - quite literally, as perhaps the most impressive and original part of Mollick's discussion is developed from a paper with colleagues from Harvard Business School: "Navigating the Jagged Technological Frontier" is a paper I've been using for some time with students, and one that I recommend anyone with an interest in AI in the world of business should read. Nonetheless, there are countless examples of AI prompts and outputs that I found myself skipping, and which reminded me that while I find AI incredibly useful as a daily tool for a wide range of uses, those daily encounters remind me that I really don't want to read bookfulls of what can all too easily become slop.
I am being more than a little mean here because in truth there is much in this book that is very worthwhile: the first part offers a good, clear primer of some of the primary issues facing AI users and developers today, such as its role in the workplace and the question of alignment - making AI serve our needs rather than become our masters or work against us. He also ends the first part of the book with what I think is a sensible and eminently useful set of what he calls "Four Rules for Co-Intelligence". Indeed, the first is so clearly stated that is one that, upon reading, I immediately determined to put into practice with my students:
You should try inviting AI to help you in everything you do, barring legal or ethical barriers. As you experiment, you may find that AI help can be satisfying, or frustraint, or useless, or unnerving. But you aren't just doing this for help alone; familiarizing yourself with AI's capabilities allows you to better understand how it can assist you - or threaten you and your job.
This, and the other three rules - be the human in the loop, treat AI like a person, and assume this is the worst AI you will ever use - are practical and focussed guides to considering AI, and what I like best about this book is that it is really about developing an attitude towards AI: pragmatic, not afraid to call out its bullshit, but also willing to use it to develop our personal and professional lives.
The Six Pillars of AI
The second part of Co-Intelligence is a mixed bag which provides six examples of how to use AI in everyday life. This is the section where, to be honest, huge swathes of generative text could have been cut out and the book would have been greatly improved. Alongside that, however, there are also plenty of genuinely useful practical examples and one section which is extremely worthwile.
Mollick breaks the second part into sections dealing with AI as a person, a creative, a coworker, a tutor, a couch and, somewhat ominously, as our future. The first five are concerned with the here and now, and for the most part they offer some useful tips from his experience of using AI on a regular basis as a professional and also in the classroom. Personally, this is one of the most useful features of this book, in that it's an adult discussion of using AI to enhance and augment his daily practice and experience. It's also full of some pretty decent advice in parts, but other sections could have been cut without much loss. Looking through my notes for the first chapter in Part II, AI as a Human, I realised I made almost no notes: part of this is because he runs through a chunk of AI and computer history with which I am very familiar (and, as a consequence, you might find useful if you haven't come across Mcirosoft's Tay chatbot experiment or want to understand Turing's Imitation Game more thoroughly). It's also a chapter that is full of an extensive chunk of AI-generated text which, while it is there for a reason, really could have been summarised. Likewise, I found the chapter on AI as a Creative not especially useful, although it makes some decent observations about the fact that while LLMs don't really know what they are doing, the disruptive factor comes from the fact that they appear better suited to creative rather than repetitive tasks.
I very much realise that I am damning with faint praise, which is why the chapter on AI as a Coworker saves this book - indeed, it is far and away the most useful part of Co-Intelligence. Drawing on research that Mollick did with Harvard Business School and Boston Consulting Group, I was aware of his contribution but this is still extremely useful as a more accessible version of that research. Mollick's background as a professor of management really helps him shine here (in a way I don't think it does when looking at creative processes), beginning with many of our own assumptions that stem from our inability to see our jobs as "bundles of tasks". Once we do that (which has had a profound effect on my own thinking), then it becomes much easier to hive off tasks to AI without the crisis of identity - and value - which often accompanies the use of AI in the workplace. The case study looked at consultants who were split into two groups: those using AI and those without. Unsurprisingly, the group with AI did exceptionally well - until the information was wrong and they they performed considerably worse. As Mollich points out:
When the AI is very good, humans have no reason to work hard and pay attention. They let the AI take over instead of using it as a tool, which can hurt human learning, skill development, and productivity. He [Dell'Acqua] calls this 'falling asleep at the wheel'."
The point is, as with so much of Co-Intelligence, lucidly made, and indeed the following chapter on AI as a Tutor also draws on considerable experience, demonstrating him at his best as a writer (even if, again, there are big chunks of interactions with ChatGPT that I really didn't need to read.
The book ends on a speculative note which isn't really the author's strength: while fair to consider what will happen if AI improves or stagnates, it has the whiff of accelerationist fantasy - both doom-mongering and celebration - that often infects writing on AI. To end on that note, however, is extremely unfair: throughout large chunks of this book I was impressed by the fact that it's one of the most even-minded, indeed sensibly adult accounts of AI that I've read in a while, one which genuinely considers how it can improve our lives as professionals (and more) while also acknowledging the many pitfalls that face us. As he remarks: AI could add a great deal to our lives - so long as we don't fall asleep at the wheel.