M.026 The Superposition
Why the AI potential of your people is dead and alive at the same time
You are reading Molekyl, finally unfinished ideas on strategy, creativity and technology. Subscribe here to get new posts in your inbox.
In 1935, Erwin Schrödinger coined one of history’s most famous scientific thought experiments: The cat that is both dead and alive at the same time.
Schrödinger’s cat was originally intended to show the absurdity of the superposition concept of quantum mechanics, if applied to a more human scale. Instead, it ended as an elegantly simple illustration of that very concept. A beloved metaphor for how a particle can exist in multiple states at once.
A question I have been pondering lately is whether the superposition concept itself could be a useful metaphor to help us better understand something else. Namely the human implications of the current state of AI.
After giving it a proper thought, I am leaning towards a yes. Let me explain how and why.
The superposition of choices
Quantum mechanics is a strand of physics I won’t say I understand much about. Still, that doesn’t stop me from being fascinated by it.
For one, the ambitions of the field can only be admired. Nothing less than attempting to understand how our universe works from studying its very smallest components.
But I am also fascinated by the beautiful weirdness of some of the theory’s predictions. Like that of the superposition, the concept that inspired Schrödinger’s cat.
In quantum mechanics, superposition refers to situations where particles exist in multiple states at once. The cat is both dead and alive, at the same time. Both states are equally real, until we measure. When we do, the superposition collapses into one of the states, while the others disappear. We open the box and look inside. The cat is no longer dead and alive. It’s either or.
Taking the leap from particles and cats to humans, we find a parallel to the quantum superposition in situations where each of us is faced with a decision or choice. In that moment, all the potential decisions we could make and their potential outcomes exist simultaneously.
I could continue in my job or leave it for something else. I could spend the evening writing, or I could watch a TV series. I could build one thing, or another. All potential outcomes equally real until I have made my choice.
But when I make a decision, either explicitly or implicitly, my superposition collapses. I decide to stay in my job, spend the night writing, or prototype one particular idea from my drawer. When I do, one outcome becomes actual, while the rest vanish.
This superposition-like state of decision situations isn’t anything new. What is new is how AI greatly changes the superpositions for some, and not for others.
Let me illustrate.
What AI does to the superposition
Until recently, my superpositions were fairly narrow, and stable year over year. The number of potential trajectories in any situation was limited by my skills and capabilities, and the potential outcomes of each trajectory were spatially close. The former because skills and capabilities accumulate slowly, and the latter because my slow speed and capacity constraints naturally limited how far I could travel down each trajectory in a given time.
For example, starting a new week, I could draft a few sections of a paper instead of another, or run one set of analyses on a dataset instead of another. The superposition of potential outcomes was narrow and the states neighbours.
The last couple of years this has changed drastically. Because of AI.
The first reason is that the number of potential states in my superpositions suddenly exploded. Potential trajectories are no longer limited by my own skills and competences. I found the bat cave, full of gear allowing me to take on challenges previously beyond my reach.
My start-of-week options are no longer limited to choosing which paper to write or analysis to run. Now they also include making music, film, design, art, software, data visualization, and so much more.
The second reason for the exploding superposition is that the efficiency- and capacity gains of AI allows me to travel further on each trajectory. The result is that the spatial distance between each of the potential outcomes in the superpositions also increases. From being close neighbours to being galaxies apart.
Now, one week with AI an individual can produce several full research papers, build functioning software, create and release music on Spotify, write and produce a full length film or two, or even build and launch a full business.
But the key word here is can. Because this hasn’t happened to the superpositions of everyone.
For those who still live their life as if nothing has happened, their superpositions are mostly the same as they ever were. Both the number of potential states and their spatial distance are the same as before. All while the superpositions of the AI-prolific has exploded, with more states and different states being further apart.
This uneven impact leads to the weird situation where every organization currently consists of people of both kinds. For some, the future is a narrow corridor of familiar options. For others, it’s an increasingly open field stretching in every direction, with genuinely different destinations visible on every horizon.
The challenge of the counterfactual
A widening superposition is the challenge of having too many roads to choose from. But since changing nothing and continuing as before is also a choice, each of us are continuously collapsing our superpositions.
You could have built that agent system last week that would have allowed you to do a lot since then, but didn’t. You could have quit your job to join the founding team of that startup that is now worth a lot, but didn’t. You could have launched that app idea half a year back, that in retrospect would have made you rich, but didn’t.
Contrary to particles, humans have feelings. We are also capable, in theory at least, of reflecting on our choices. Combined, these two features lead to a second challenge from an exploding superposition beyond just the expanding menu of choices: the challenge of the counterfactuals.
Counterfactual refers to what would have happened if things had been different. In this case, if you had chosen differently. Taken a different road than the one you took.
If the superposition is the challenge of choice, the counterfactual is the challenge of living with the choice you took, knowing that things could have been different.
For a narrow superposition where states are close neighbours, the counterfactuals are relatively mild. What we gave up by choosing something was a relatively minor variation of the actual outcome. The road not taken looked a lot like the road we took. The regret, if any, is therefore also relatively small.
When the superposition widens, this changes. For the very AI prolific, who now can do so much more than before, every collapse is comparably much more violent. Choosing something doesn’t just forfeit a slight variation of one trajectory. One single decision might suddenly destroy genuinely different futures.
For those who are free to choose, like the indie builders who ship a new business every week, or the Anthropic engineers who build software over a weekend that shakes up entire industries, the counterfactual is mostly regret. The freedom to choose makes the weight of the counterfactual pure and personal. They picked one trajectory, and have to carry the ghost of the counterfactual outcomes.
What if I had taken that offer from OpenAI in 2019 instead of joining Salesforce? What if I had quit my job to build that AI startup I dreamt up 2 years ago, instead of climbing the corporate ladder at my current employer? With AI, such regrets can be big and violent, as states can be galaxies apart.
The regret is pure and personal when we are free to choose among our alternatives. But in any practical terms, most of us don’t operate with full freedom. We work inside organizations where we need to adhere to leaders, rules, compliance, hierarchy, norms, organizational habits, and so on.
All of this is key to making organizations function, but it also serves as an outside force that blocks trajectories from people’s superpositions.
For the most AI prolific, the implication is that many of the most interesting states in practice become impossible to choose. Clearly promising trajectories that someone or something prevents you from choosing. Without that someone or something necessarily knowing that they do.
When decision alternatives are blocked by others, it’s no longer the regret of having chosen wrong in the decision situation that is the issue. It’s more the frustration of not being allowed to choose at all. Feeling that genuinely different futures sit right there within reach, but then having someone or something else sweep them away. You might never have taken any of the roads anyway, but that doesn’t really matter for the feeling.
For leaders, the tricky part is that the people who feel this frustration most acutely are likely the ones with the widest superpositions. The most capable, most AI-prolific people in the organization. Their superpositions are suddenly vast in theory, but much more narrow in practice due to organizational blockages. Something that can create frustration for the people involved, and destroy potential value for the organization they work in.
That feeling, explained
Combined, I think that the changing superposition of AI might explain why that feeling I tried to articulate in this earlier piece hits some people like a freight train, while others are largely untouched.
If you haven’t spent much time with AI, your superposition is narrow. The states are close together. Choosing is easy enough, and what you give up by choosing is small. You look around and see a world that looks roughly the same as it did two years ago. You feel fine.
If you have spent serious time with AI, and felt what it can do when you bring real problems and real expertise, your superposition has rapidly expanded. Looking ahead the options are suddenly so far apart that choosing feels harder than before. And when you choose, you increasingly feel that you simultaneously leave genuinely different futures behind.
In Silicon Valley, it seems that many are feeling this in their bones. Exemplified by this recent tweet describing the current sentiment in the bay area around AI. It reads almost like a field report from inside an exploding superposition. People demanding higher salaries and switching jobs. People being paralyzed by the question of whether they’re climbing the wrong ladder. Middle managers who are frozen and frustrated because they can see the roads, but can’t reach them.
“Am I in the right place? Should I move? What if I had applied to OpenAI back in 2019? Or launched that startup idea? Is there time still left?”. All this is the sound of expanding superpositions, and their human implications.
I do, however, think it would be a mistake to believe that this is limited to Silicon Valley. Every organization now has people with a greater variety of superpositions than before, gathered under the same roof. Some have a self-proclaimed AI maniac at the top, like the Norwegian Oil Fund, that see exploded superpositions everywhere. Others have leaders who see the world much like they did before AI.
For those in the latter category, I think it’s especially important to be aware of how AI has expanded the superposition of some of their employees and peers. If not, it’s easy to inadvertently frustrate the people making up your biggest AI potential, by continuously blocking the many new and interesting roads that AI opens up for them.
At best, these people represent a latent, unrealized potential for your organization. At worst, the very frustration this creates might make people who never thought about leaving, start looking for opportunities elsewhere. Where the bounds on their superpositions are fewer.
And maybe that is the true insight here. That the AI potential of your people is both dead and alive at the same time. And that leaders have a role in making sure this potential is alive when the box is finally opened.