Fire Did Not Ask Permission

When our ancestors gained control of fire, they did not vote on it. The technology arrived, distributed unevenly, and everything downstream — diet, shelter, social structure, the length of human days — reorganized around it. Fire cooked food, which changed gut length and jaw structure over generations. Fire extended productive hours, which changed what people did with time. Fire required fuel, which influenced where groups could live. A single tool rewrote the terms of human existence. AI is a different kind of tool in important ways. It is cognitive rather than physical. It scales digitally rather than requiring physical resources to replicate. Its effects are immediate rather than playing out over generations. But the structural parallel is real: we are dealing with a capability that is general enough to reorganize downstream conditions across multiple domains simultaneously.

The Creation Side

The case for AI as creative and generative is not hard to make. AI systems are already accelerating drug discovery by modeling protein folding and molecular interaction at scales no human team could match. They are producing code that extends what individual developers can build in a given timeframe. They are making translation, summarization, and information retrieval faster and cheaper, which in principle democratizes access to knowledge. In science particularly, the effects are beginning to accumulate. Research published through collaborations between DeepMind and academic partners on protein structure prediction has already influenced thousands of downstream research projects in biology and medicine. Tools that would have required large, well-funded laboratories to build are now accessible to small teams in countries with limited research infrastructure. That is a genuine creation — new capability distributed more broadly than the old capability was.

The Destruction Side

Fire burned things too. That was not incidental — it was part of the value. You burned the field to make it fertile. You burned the wood to make the warmth. The destruction that AI introduces is economic first. The categories of work most vulnerable are not the ones that require the least skill in any simple sense, but the ones that require skill that can be replicated at low cost: information retrieval, first-draft content generation, routine analysis, pattern-based coding, standard legal research, basic financial modeling. These are skills that took years to develop and are now being reproduced by systems that improve continuously and scale without friction. This is not hypothetical displacement. It is documented. A study by researchers at MIT and Boston University examining the introduction of AI tools into professional services firms found measurable reductions in entry-level hiring within two years of adoption — not because firms were struggling, but because the work that had previously required junior staff could now be handled by AI with senior oversight.

A Tangent on Who Controls the Tool

Fire, in principle, was available to anyone who could maintain it. In practice, control of fire was a form of social power throughout early human history. The parallel question for AI is about who controls the systems and the infrastructure they run on. The organizations building foundational AI systems are few, well-resourced, and located in a small number of countries. Access to the most capable systems is gated by cost, internet infrastructure, and sometimes by deliberate policy choice. The generative potential is real, but the distribution of that potential is not automatic or guaranteed. The utopian scenario where AI democratizes capability requires deliberate choices about access and pricing that are not economically guaranteed.

The Tool That Changes the Toolmaker

What makes fire a useful analogy beyond the creation-destruction frame is what it did to the toolmakers themselves. Using fire changed human cognition over long time scales — cooked food provided more energy for brain development, longer social evenings created more time for language and storytelling. The tool changed the users. AI is likely to do this too, on shorter time scales and through different mechanisms. Skills that are offloaded to AI systems tend to atrophy in the humans who stop practicing them. Navigation ability declined as GPS use increased. Arithmetic fluency declined as calculators became universal. The open question is which cognitive capacities will atrophy as AI takes on more cognitive work — and which will be freed to develop further.

Calibrated Response to Both Sides

The honest position is that AI brings both creation and destruction, as fire did, and that the ratio depends significantly on choices being made now about deployment, access, governance, and education. Tools do not have intentions. Their effects are shaped by the social, economic, and institutional contexts into which they are introduced. The people and institutions that navigate this well are those that resist both the utopian and dystopian framings — neither pretending the disruption will be painless nor treating the benefits as illusory. Fire was worth it. AI probably is too. But "worth it" does not mean the costs will be distributed evenly, and pretending otherwise has never been accurate about any major technology transition.