The Mentor | #67 | Return
Reflections on work, leadership, developing careers and living a bit more deliberately.
During my eighty-seven years I have witnessed a whole succession of technological revolutions. But none of them has done away with the need for character in the individual or the ability to think.
-Bernard Baruch
The Mentor hiatus was helpful. I’ve decided to restart with a new format (less links) and new schedule (whenever I have something to say). Will strive to keep it succinct and useful, especially for those looking to develop themselves and their careers.
AI compelled me to write again. Maybe it’s because AI-generated writing is so bad. Or maybe it’s just because it’s weird.
It’s true. AI is weird. It can write a song that sounds right but cannot actually be played. It can do analysis I once thought required distinctly human judgment. It can be both overhyped in some places and underestimated in others. It will create productivity. It will create new jobs. It will disrupt some existing ones. All of those things can be true at once.
As I have spent more time with AI, three lessons keep ruminating in my head.
***
I am your very average campfire guitarist. Which means that, if I know the words, I can usually pick up a guitar, figure out the three (sometimes five) chords to a song, and play it. This works best with country and folk music, reasonably well with rock, and not at all with rap or R&B.
I recently received an AI-generated song. When I listened to it, it sounded good. Correct, even. But when I tried to transcribe the chords, parts of it were easy and other parts were literally impossible. I could hear the words. I knew the chords. But nothing would quite fit the actual flow of the song. The chords had to change, even though the tone did not. It was completely perplexing.
The song was plausible. It just was not playable.
That experience helped me understand something I recently heard Jensen Huang say about AI: separate the job from the task. The task of a campfire guitarist is to convert verses into chords. AI had done something like that. But the job is to play the song around the campfire…to make it work in the real world, with actual people, in an actual moment. That requires more than producing something that sounds roughly right.
Someone later pointed out that my attempt to rewrite the song was a real-world example of “evals.” AI had given me a useful starting point. But I still had to decide whether it worked.
Lesson 1: Separate the job from the task. AI will be very good at many tasks. Your value will come from understanding the whole job and knowing whether the output actually works in the real world.
***
There is a pre-you and a post-you once you really confront AI.
Most people think they have used AI because they have had a chat or search experience in an app. That is interesting. It is exciting. But it is still mostly the pre-you version. It does not completely change your worldview about what is possible.
My post-you moment came when I saw AI integrate multiple data sources, digest a large set of documents, and then write a strategy and organizational design based on all of the inputs. It was not just fast. It was right. It was doing a kind of critical thinking that I had not believed was possible from AI.
I did not leave thinking AI could answer questions faster. I left thinking it could participate in work I had previously considered distinctly human.
Lesson 2: There is a pre-you and a post-you with AI. Find a way to get to post-you soon. It will expand your view of what is possible and perhaps make the whole thing feel a little less threatening.
***
For most people and companies, AI is not primarily a technology problem. It is an operating problem. By operating problem, I mean: how exactly do you make AI part of what you do, rather than one more thing you have to remember to use?
In almost every conversation I have about AI, the question is no longer, “Is this real?” It is, “How do we actually use it without creating chaos?” That is why AI tends to evoke two related emotions: FOMO and FOMU.
FOMO is the fear of missing out: are others doing something that I am not, and therefore am I falling behind?
FOMU is the fear of messing up: if I do something and it does not work, will I become the problem or the reason things did not go as planned?
The harder part is that no framework can do the work for you. Each person and each company has to dive in, build something, and figure out where AI actually fits.
Lesson 3: There is no substitute for doing the work yourself.
***
These 3 lessons of AI are not that human judgment matters less. It is that it matters more.
The people and companies that benefit most will not be the ones who talk about AI the most. They will be the ones who learn where it helps, experience enough of it to see what is possible, and then do the unglamorous work of changing how they operate.
I love the distinction between job and task. It really points to where we, as humans, need to focus understanding our true value. AI is clearly pushing us out of our comfort zones, encouraging us to move away from routine tasks and toward more meaningful, higher level contributions.
Excellent post. I can give you another example from my own life. Many moons ago, I was a creative writing undergrad — before IBM and before my previous career. Recently, I’ve reconnected with that part of myself.
There is an AI I use that makes for an excellent editor. It is great at that task, but not at the job itself — or in your example, the art. It can absolutely help with what a human editor does, and in some cases I’d argue it performs better than many. But it cannot create the story that truly connects with a reader. That still requires the human element.
Again, great post. Keep them coming — I know you will.