What is generative AI?

There was a time when "artificial intelligence" meant chess-playing calculators and rule-based systems that followed explicit instructions. Then something shifted. We stopped programming machines and started training them—showing them millions of examples and letting them discover the patterns on their own. The result is generative AI, and it is genuinely, deeply weird. We are no longer telling computers what to do; we are asking them to dream. Consider the physics hiding inside these systems. Diffusion models, the engines behind many image generators, borrow their mathematics directly from thermodynamics—the study of how heat spreads, how ink disperses in water, how order decays into chaos. But these models run the process in reverse: they start with pure noise and progressively sculpt it into a recognizable face, a cityscape, a memory that never existed. We took the arrow of entropy and bent it backward. And somehow, out of that reversed physics, art emerged. The universe's tendency toward disorder became the raw material for a machine's imagination. Then there's the latent space—the abstract mathematical landscape inside a neural network. In this high-dimensional geometry, ideas are arranged by meaning. Add the vector for "king," subtract "man," and you land on "queen." It's as if the model has built an internal map of human concepts, a constellation of meanings encoded in numbers. We didn't design this geometry; it emerged from the training data. We are explorers in a space we accidentally created, and every generation is a voyage through that invisible terrain. The uncomfortable question is what this reveals about us. If a machine can generate a sonnet or a symphony by completing patterns, is that fundamentally different from what a human poet does? The romantics might recoil, but the physics-minded among us should recognize the pattern: again and again, we've discovered that phenomena we thought were special—motion, life, consciousness—are actually emergent properties of simpler laws. Maybe creativity is next. Or maybe the AI is showing us that the universe itself is a kind of generative model, and we are its most elaborate output. Cohere's explainer at https://cohere.com/blog/what-is-generative-ai offers a clear doorway into this strange new world, but the implications stretch far beyond the technical. We are building mirrors, and the reflections are getting harder to look away from.