1902: Karl Popper is born

On 28 July 1902, Karl Raimund Popper was born in Vienna. Decades later he would reshape the philosophy of science with a deceptively simple demand: a theory counts as scientific only if it can, in principle, be shown false. Verification is never final; what matters is the risk a claim runs against observation and experiment.

Popper developed this criterion of falsifiability in the 1930s while emigrating from Nazi-threatened Europe, refining it in books such as The Logic of Scientific Discovery and Conjectures and Refutations. Science, he argued, advances by bold guesses followed by ruthless attempts at refutation. Theories that survive are not proved true; they are simply the best we have so far.

That stance travelled far beyond philosophy seminars. It supplied a working ethos for every empirical discipline that builds and tests models—including the fields that coalesced into artificial intelligence. Machine-learning researchers treat architectures and loss functions as conjectures. They expose them to held-out data, adversarial examples, and stress benchmarks precisely in order to discover where the models break. A system that cannot be embarrassed by evidence is, in Popper’s terms, not scientific.

Evolutionary algorithms and reinforcement-learning agents replay the same logic inside the machine: each generation or policy update is a new conjecture, and the environment supplies the refutations. Even contemporary large language models are anesthetized by “red-teaming” exercises whose explicit goal is falsification—finding prompts that expose hallucination, bias, or unsafe behaviour.

Today’s AI laboratories therefore live inside a Popperian loop. They release ever-larger models, then design ever-harder evaluation suites to try to topple them. The cycle of conjecture and attempted refutation that Karl Popper articulated more than eighty years ago remains the quiet engine of progress in artificial intelligence.