How do we plan for the future when technology changes this fast?
A few years ago I read Smarter Faster Better by Charles Duhigg. There’s a chapter about forecasting that I keep coming back to.
At the most basic level, we predict things all the time. If a child reaches for a hot pan, we stop them because we already know what is likely to happen next — they get burned. We can predict those next few seconds because we already understand the pattern.
Longer-term prediction and professional intuition work similarly. A firefighter, doctor or pilot sees thousands of situations over time. Years of experience teach them which signals matter and what usually follows.
Humanity has been through huge technological shifts before: electricity, factories, cars, the internet, smartphones. They changed jobs, businesses and everyday life.
Today we’re living through another one, with AI changing many of the rules we’ve become used to.
You learn what a system can or can’t do, and a few months later some of those assumptions may already be outdated. You build around a product gap and a platform closes it with a new release. A company spends months implementing software while better options appear. A profession may still exist, while the skills and tasks that create the most value within it can shift much faster.
The next change can arrive before we’ve adapted to the previous one, making long-term planning harder.
And people still have to choose a profession, build a company, decide what to teach their kids, which skills are worth years of effort, and what is worth creating today.
Predictions about AI are everywhere, and they are extremely polarized.
AGI, humanoid robots, abundance, universal high income.
Mass displacement, disappearing professions, declining human skills, loss of control, existential risk.
That level of uncertainty is exhausting. Add FOMO, information overload, attention economy. It’s hard not to wonder where all of this is going.. and that’s before adding wars and political instability.
A few years ago, during the hardest period of my life, one of the most useful things I learned was to deal with uncertainty by turning fear into a scenario.
If X happens, what will I do? ✔️
A plan (or better, a few) makes uncertainty easier to handle. Philip Tetlock’s work on forecasting makes a lot of sense in this context. Good forecasters keep several outcomes in mind and update them as new information comes in.
Keeping up can take a ridiculous amount of attention. What helps me is keeping more than one scenario in mind, staying close to what is actually changing, learning continuously, experimenting, testing things myself, stepping outside my information bubble from time to time.
My forecast for the future:
Somewhere between Elon’s universal high income and superintelligence killing us all.
There’s still a lot we can’t predict in between…
I have a dependency problem.
I managed to break free from one particularly aggressive one. It cost me a lot of time and money. It was an IT outsourcing company.
That experience taught me to recognize and assess the risk of similar dependencies before getting involved, to be more careful about who I trust and what I hand over, to see the bigger picture, do things differently, and get on a train that’s now racing across the world at insane speed.
If someone confidently tells you where this AI train is headed and what the future holds, I wouldn’t be so quick to believe any of it. No one can say for certain where it will take us. We don’t have enough experience with this technology yet to reliably assess which future scenarios are most likely.
In today’s world, complete independence already seems almost impossible. But some things deserve special attention.
I’m taking back my memory, my data, my intentions and the leverage they give others to manipulate me. I’m building a system where I keep what’s most valuable.
Few people will understand what I’m talking about today. Sooner or later, many users of today’s human-facing internet will want to do the same. By then, it will be harder.
Time to cancel a lot of subscriptions.
Not all of them, unfortunately.