The Old Road Is Still Crowded

The spark for this rabbit hole was Sam Altman’s recent essay, _Industrial Policy for the Intelligence Age_. It is worth reading directly, so I am not going to rehash it here. What mattered for me was the thought it triggered: sometimes the real barrier to a new technology is not the technology itself, but the old world it has to share space with.

Echoes, Divergences, and the Uses of Analogy

There are moments in history when the next machine is not blocked by its own weakness, but by the lingering shape of the world around it.

I keep wondering whether AI may be in one of those moments now.

Take the self-driving car thought experiment. It is at least plausible that we could create something much closer to an efficient autonomous taxi system if the roads, rules, and traffic were designed mostly for machine drivers. Not perfect. Not universal. But materially better.

The problem is that machine judgment does not enter a vacuum. It enters a road already crowded with human impulse, human ambiguity, human pride, human fatigue, human improvisation, and the occasional sacred act of complete nonsense.

That possibility feels larger than cars.

In many domains, the machine may not be failing because it is too weak. It may be failing because it is being forced to share a lane with institutions, incentives, and behaviors shaped for an earlier operating system.

The Horse, the Car, and the Long Middle

The automobile did not simply appear one morning and send the horse into retirement by lunch.

The transition took decades.

The car became credible before it became ordinary. It improved, roads improved, manufacturing improved, prices improved, standards improved, and public expectations changed. In some places, the shift happened quickly enough to feel dramatic. In others, the old and the new lived side by side for years. Even after cars became common in cities, horses remained in freight, farming, and edge cases where the older system still worked well enough.

That long overlap matters.

We often tell the story of technological change as if one thing defeats another in a clean duel. But that is usually not how it happens. More often, society lives in the messy middle. The old system declines unevenly. The new system arrives unevenly. Infrastructure lags. Rules lag. Habits lag. Training lags. Politics lag. Meaning lags.

And human beings, being human, continue trying to solve tomorrow’s problems with yesterday’s defaults.

Why Analogy Still Matters

None of this proves that AI will follow the same path.

Analogy is not proof.

But analogies that work are still useful because they help illuminate what is possible, what is plausible, and what kinds of transition dynamics we should watch for.

Human critical thinking would stop cold for most of us if we refused to rely on analogous experience. We do not survive novelty by pretending we have seen it before, exactly. We survive it by noticing what it rhymes with.

Analogy is one of the oldest tools human beings have. It is how we drag a little light from the known into the unknown—not to guarantee, not to predict, not to imprison the future inside the past, but to orient.

Past performance does not predict the future.

But it does inform us.

It gives us shapes to compare, warning signs to watch, transition patterns to respect, and false certainties to distrust.

The mistake is not using analogy. The mistake is worshipping it. The disciplined use of analogy asks a better question: not “Will this happen the same way again?” but “What does this older transition help us notice?”

What the Parallel Suggests About AI

That is why the horse and the automobile matter here.

Not because AI is a car.

Not because workers are horses.

Not because history runs on rails.

But because the road itself had to change before the machine could become ordinary.

That may be true for AI as well.

We keep asking whether the machine is ready when, sometimes, the truer question is whether the surrounding system is. Many technologies do not fail because they are incapable. They fail because they are forced to operate inside environments full of mixed signals, conflicting incentives, legal ambiguity, legacy structures, bad handoffs, and human edge cases that no one has chosen to redesign.

Seen this way, the first victory condition for AI may not be full replacement.

It may be limited corridors where the environment is structured enough for the machine to perform well.

It may be hybrid institutions.

It may be workflows where humans and machines do not so much compete as take turns holding the wheel.

It may be carefully designed transition zones where we let the new system prove itself without pretending the whole world is already ready.

That is not failure. That is how transitions often work.

The Mixed-Mode Era

This is the part people tend to underestimate.

The overlap period is expensive. It is messy. It is politically annoying. It frustrates purists on both sides. From a distance, it can look like hesitation. In practice, though, the mixed-mode era is often where the real work happens.

That is where organizations learn what must change around the technology, not just inside it.

That is where standards emerge.

That is where accountability gets negotiated.

That is where workers figure out which uses actually help and which ones merely shift burden.

That is where society slowly learns the difference between a powerful demonstration and a durable system.

If AI follows this pattern, then the future may arrive in districts, corridors, and partial stacks before it arrives everywhere. The machine may be locally ready long before society is broadly ready to reorganize around it.

And that may be enough to matter.

Closing

The horse did not vanish in a single act of technological triumph.

The road changed beneath it.

The city changed around it.

The economy changed above it.

And only then did the old rhythm begin to fade.

AI may ask the same of us.

Sometimes the future is not late because the new thing is weak.

Sometimes the future is late because the old road is still crowded.

And sometimes the past does not hand us a map of the future.

But it does leave cairns on the trail.