This content is locked. Please login or become a member.
Deciding Where to Focus
We have a lot of early evidence that this is going to be a big deal for work. Some subjective evidence, but also a lot of objective evidence. So there’s now multiple studies across fields ranging from consulting to legal, to marketing, to programming, suggesting 20 to 80 percent performance improvements across a wide range of tasks for people who use AI versus don’t. When steam power was put into a factory in the early 1800’s, it improved performance by about 18 to 22 percent. So these are really large numbers, like, very large. And so there’s a lot to think about about how that fits in and about all the other things AI does. It’s seemingly creative, out innovates most people.
One example, by the way, is that I assigned my students the task this semester of automating their jobs with AI. I have MBA students from all over the world, all kinds of professions. The task was you go to your next job interview and you slide over the
prompt or the GPT, and you say, my job is done. Please give me a raise. And I had students who were navy pilots and hip hop promoters and lots of private equity and venture capitalists and consultants and real estate investors. And all of them came up with tools that automated parts of their job. I would never have thought that this could be used to file a Navy flight plan or that it would be really good at being able to do a very particular kind of analysis that people do when doing real estate investment. My students were able to figure that out because they knew the industry well. What you know well is where you wanna use it because you understand what it’s good and what’s bad at out there.
Assigning Tasks
Ways of working with AI depend on your circumstances, your experience. You could divide the world into “just me” tasks, things that either the AI can’t do or is important for me as a human to do them. So these are things that matter to me, that I wanna be in the loop for. And so sometimes AI can do those tasks, sometimes it can’t. That will probably be a shrinking number of tasks as we hand more and more stuff off to AI as it gets better and better at more things.
Then we have the idea of delegated tasks. It’s tasks where we work with the AI on things. There’s actually two ways of delegating tasks to the AI. One way of delegating tasks is what we call the Centaur Model, like the half-person half-horse, where there’s a clear dividing line. So it could be, I like to do analysis. I do not like to write emails. The AI writes the emails. I do the analysis. So divide it up, delegate tasks. Then there’s the more complicated view of the Cyborg, blended machine and human.
In my most recent book, Co-Intelligence, I wrote it as a Cyborg. There’s almost no AI writing in the book, that’s acknowledged as AI writing, but I couldn’t have written it without AI. When I couldn’t finish the sentence, it gave me twenty versions to finish the sentence. It helped me think about analogies. It helped me summarize the papers that I kept track of. I had it act as a reader and read over every chapter from different perspectives and give me feedback on it. So the AI made the writing possible even though it wasn’t the writing there. That was cyborg work.
Outside of these delegated tasks, we then move on to the idea of automated tasks, of agents. And increasingly, that’s where AI is heading, is this idea of agents where we delegate something to the AI and it does it itself and just reports back to us on its own. Having that framework of “just me” tasks, of the two kinds of delegated tasks, of Cyborg and Centaur, and of automated tasks where the AI just does the work on its own, you’ll find yourself naturally becoming a cyborg. Because you’ll naturally say, Oh, yeah, it could do this. Why would I bother doing it myself? Or this looks like an AI task, but it’s actually something that should be done by me.