“We all make choices, but in the end, our choices make us.” - Andrew Ryan
Homogeneity
Stop me if you've heard this one before.
A CEO walks into a bar and orders champagne, declaring that he has saved his company millions by firing all his staff and replacing them with AI. The board members cheer, as stock prices go through the roof. Capitalism has finally left behind the tyrannical reign of the work force, selfishly demanding salaries and time off.
Another CEO walks into the same bar and declares the same, more cheers.
A third CEO walks in and starts to declare the same thing, this time there are some confused looks around the room.
Then a fourth and a fifth.
By this point some of the CEO's have converged like a gaggle of geese.
'Well my AI told me this was the best place in town' announces one of the CEO's.
'Mine too' shouts another as their champagne sloshes in the glass.
'Wait so did mine... you don't think that we are all using the same models?'
'Well I only use the best of the best, you wouldn't believe how much it costs me, thats actually why I had to fire my staff' they say, laughing with an insidious grin plastered on their face.
'O me too' says another, raising his now empty glass to get a refill.
Wait if we are all using the same models, and we are all getting the same advice, and the same output... how will my products be any different than yours...' ponders the last CEO.
This is the homogeneity paradox, a business thinks that using AI will give it the ability to innovate and produce unique products at a faster rate, when in fact the artifacts produced become more and more like its competition.
Mental Models
The challenge to over come is a cost centre mentality. Our industries have become obsessed with cutting cost, on paper this sounds purposeful and obvious. If we can save money our profit margin goes up, its a clear signal on how to add share holder value.
But that alone isn't enough, we've also become obsessed with data first decision making. Combine these two concepts and firing half your staff and replacing them with a machine is as obvious a conclusion as 1 + 1 = more money.
But the secret sauce that makes one company a success and another one a failure can't be measured, it isn't created by the output of the work force, it is created by the effort of the work force.
As you ponder and process a problem, as you hit your head against it, struggle and devour it, that is the secret sauce that is the value add. It's ineffable, but the knowledge you gain, the questions that you, your team and your company explore is what creates the USP. The output, the artifacts that are produced, that is the side effect of that value add process.
The idea that we can replace that, that is a fallacy that compounds the Homogeneity paradox. After all effort is a feature, not a bug.
I heard an engineer explaining how he couldn't keep up with his work force of AI models, their output exceeded his grasp. But what that really means is that he is outsourcing his processing to those models, he no longer understands the details with as much depth and clarity as he would have before using AI to create these artifacts.
Critics would cry that of course he does, now he has even more time to understand the architecture, the design and the core problem of the issue. But I've run teams of engineers and I can tell you that the only time I've known the complete detail of a solution is when I'm the one creating that solution. When others in my teams solve an issue, I can probe it, I can understand it, but I will never know it as intimately as when my hands were the ones that crafted it.
That is fine when dealing with humans, because you trust that they do understand the problem with the level of intimacy that it requires. You understand that you can go back to them and follow up on details, days or weeks or months later and together build on top of those foundations. Now compare that to an AI, as soon as the problem is out of it's context window it has to start again, it will change it's opinion between two different prompts, thats not a worker you can build trust with, thats a slow horse you occasionally want to shoot.
The more we outsource our understanding the less unique our impact can be, the more we ask AI to answer our questions and the more we parrot their response, the more homogenous our thought processes become. Homogenous with the average of a statistical data model, homogenous with every scrap of data thats ever been scraped from the internet, homogenous with every dumb idea thats been posted on the internet and lets face facts, there are a lot of dumb ideas on the internet.
The most impactful companies are not the ones that do the same as everyone else, they are the companies that at one point of another cut their own unique path, and that is something an AI by its very nature will never be able to do.
It is the uniqueness between each of us, the friction between our differing points of view, the effort of understanding and collaborating that creates miraculously, inspiring and impactful artifacts, whether thats art, software or a hand knitted tea cosy. That is what really adds value, that is the secret sauce that every company desires, that is what helps change the world for the better, your unique point of view and how you work with others to make it even better.
"You are a unique snowflake and thats awesome"
And as Andrew Ryan said We all make choices, but in the end, our choices make us. What will your choice be? How much of the secret value add, that you bring, are you willing to outsource, because as far as I'm concerned the effort is the feature it is not a bug.
My Promise To You
And so here is my promise to you, and more importantly to myself. My blogs, will be written by me, not an AI. The writing style for better or worse is mine and mine alone, it is unique, imperfect and will hopefully improve over time.
I will use this space to help hone my own thoughts, to process, to improve. And as a side effect, I will create these artifacts, these posts that are created by a human for humans. Because the effort involved in producing them is a feature, not a bug.