I can't say that I was ever particularly good at performing the duties of my job.
Towards the end of my career some might say I was useless. I can accept that.
But that's because I am an outlier.
In the 1990s this wasn't always considered a bad thing. In certain circles it was positively useful to have one around. You had the process people, the account people, the finance people and the ones who could tell you why something couldn't be done, usually with a spreadsheet.
And then occasionally there was somebody who looked at the same thing and said, ok, but what if we did this?
Sometimes it was stupid and sometimes it was really bloody stupid. But there were times when it changed everything.
The outlier delivered asymmetric advantage more times than he didn't. Wait, what's an asymmetric advantage? We’re in a boxing match and I'm the much smaller guy - we’re both using gloves but I’ve designed a gun-toting version for small hands only. It’s about being so different that your competition's strengths become irrelevant.
And the outlier delivered not because he was sitting there trying to be original. At least I wasn't.
In fact, quite the opposite. I found that the best stuff usually happened when I’d forgotten to be careful.
I became consumed in the problem, or the game of it, surrounded by people who were also playing, and something appeared. An idea, an angle, a way through that wasn't sitting neatly inside the brief.
You couldn't always explain where it came from and that was partly the point.
Then the data scientists arrived. Not literally in one afternoon obviously, although it sometimes felt like that in the office.
By the millennium the game was changing. Measurement became more sophisticated. Process improvement became a discipline. Risk could increasingly be identified, quantified and, theoretically, removed.
All sensible. Very difficult to argue against ‘sensible’ - like ‘plausible’ these days.
The problem for all those data constricted folks is that the outlier is not particularly sensible.
His contribution is often difficult to measure, almost impossible to repeat and, worse, inconvenient. You can’t reliably put it in a process chart with predictable outputs.
So gradually the emphasis changed. Repeatability became more valuable than exception, discovery was contained within the confines of data and evidence became more valuable than your instincts. Therein lies the problem of modern judgement.
Again, none of this is a complaint.
Things change. The interesting bit is what happens to the person who spent years operating slightly outside the machine when the machine eventually becomes the dominant environment.
At first he becomes inconvenient to the system. Eventually he becomes inconvenient to himself.
And that is where it gets more interesting.
THE MAP
In 1928 J.J. van der Leeuw published The Conquest of Illusion. He was suspicious of the intellect's habit of breaking reality into categories, concepts and neat little things you could understand. Let’s call it 'the map.'
His point was that life isn't actually like that. Reality is organic, continuous, more like a plant growing than a filing cabinet. You can map it all, chop it into tidy blocks for containment and processing, but a real garden has that annoying habit of growing somewhere the map never anticipated.
As an outlier, I like that a lot.
Around the same time, though, a lot of the opposite was going on.
A German mathematician, David Hilbert, was asking whether every mathematical truth could be settled just by following a fixed procedure. No insight, just steps. A few years later Turing answered him. That answer became the blueprint for the computer.
Meanwhile Edward Bernays was writing openly about engineering public opinion.
Different people. Different problems. Different intentions. I'm not pretending they were in a room together deciding the future of humanity. But look at the direction of travel.
Classification. Computation. Prediction. The engineering of acceptable outcomes.
The map got more and more powerful. And it is powerful. That's the point. The mistake is to start believing it's the garden.
Being able to survey both the map and the garden is what kept me in a job.
THEN AI TURNED UP
Modern AI gets sold as a machine for breakthrough thinking. The thing that'll make you faster, smarter, more insightful. Your own little army of analysts sitting quietly in a server farm somewhere. And, being completely honest, it's really useful.
I use it. Research, data crunching, building digital frameworks and prototyping, project management, supplier comms. I'd be an idiot not to.
But I have a problem with the idea that this is where asymmetric judgement comes from.
Large language models work by learning the patterns in what already exists and producing the most probable continuation. Then human feedback nudges the outputs towards answers people find useful, safe and acceptable. That's not me saying AI can only produce average ideas, it clearly can't. It can recombine things in ways that genuinely surprise me sometimes.
But there's still a basic problem. It's working from what already exists. And we're all feeding it remarkably similar material, asking it remarkably similar questions, and being remarkably impressed when it hands us remarkably polished answers. All very ‘plausible’ and convenient.
So here's the danger. And it's not that AI makes you stupid (cognitive debt aside) - that's just too easy.
The danger is that it makes acceptable thinking so cheap and so available that you slowly stop noticing the difference between an answer that sounds right and something you actually know.
I see it constantly. A question appears. And before you've properly sat with it, before it's even had the chance to become yours, you type it into the machine. The machine answers instantly, and because the answer is plausible, your own half-formed thought suddenly feels unreliable. So you check it and then you check the check. And somewhere in that process, the tool you reached for to feel more certain starts becoming another reason not to trust the answer that was already forming.
That's far more interesting to me than whether AI is clever.
Because the thing being lost is the separate space where your own thinking forms before the machine fills it. And that space doesn't survive by accident. It has to be kept deliberately, like a clean room in the semiconductor industry, sealed off from all the polished, probable noise long enough for something actually yours to form.
HOW THE OUTLIER IN YOU DISAPPEARS
It doesn't happen all at once.
Nobody holds a ceremony where you surrender your judgement and somebody gives you a login. Although that sounds like quite a good film to me.
No, it happens in small changes. You discover which answers travel well through a room you’re in. Which opinions need too much explanation. Which aspects of the things you see have become awkward.
You then learn to translate what you know into something measurable, defensible and repeatable.
Sometimes that’s useful and you can advance your career that way. I did plenty of that.
But, unfortunately, something gets lost every time. And after a while the translated version becomes the one you keep using.
This doesn't just happen at work you know. You learn how a successful man behaves, what a responsible husband does, what a father's meant to be and what he's meant to provide.
What somebody at your age ought to want and what constitutes a sensible next move. And then without noticing, you build a perfectly reasonable life around a set of instructions that may or may not have anything to do with you.
And maybe one day the thing stops working and is not necessarily catastrophic. You just start to get this nagging sensation that you are playing a game you’re no longer fully connected to.
Here’s the thing. At this stage, even if you become aware of your separation, you can’t leave the game because that could look very ‘irrational’.
And this is where clever people can become spectacularly trapped because they’ve built an intellect to justify almost anything.
Stay, wait, push harder. Take six months off. Get another job. Sell the business or don't sell the business. You can build an argument for all of them.
So you keep thinking and thinking and thinking until the thinking itself becomes the blockage.
At some point you have to get out of the machinery. Step off the map my friend. Just long enough to see whether something else is there.
¹ Hilbert, D., & Ackermann, W. Grundzüge der theoretischen Logik (1928). Turing, A.M. On Computable Numbers. Proceedings of the London Mathematical Society (1936).
² Bernays, E. Propaganda. Horace Liveright (1928).
³ Van der Leeuw, J.J. The Conquest of Illusion. Alfred A. Knopf (1928).
