A neural network begins its life in chaos. Before any training occurs, the weights that connect its neurons are set to random values, typically drawn from a carefully calibrated distribution. The network, at this point, knows nothing. It produces random outputs for any input...

In the early 1970s, Bob Metcalfe and David Boggs were working on what would become Ethernet at Xerox PARC. They faced a problem that would become fundamental to networked computing: what happens when two devices try to send data at the same time on a shared wire? The signals collide, and both transmissions are corrupted. The obvious solution is to have each device wait and try again. But if both devices wait the same amount of time, they'll collide again. And again...

Sometime around 1946, the mathematician Stanislaw Ulam was recovering from an illness and playing solitaire. He tried to calculate the probability of winning a particular layout using combinatorial methods, the kind of exhaustive counting that mathematicians prefer. The numbers were intractable. There were too many possible arrangements, too many branching paths. So he tried something different...

On December 1, 1969, the United States held its first draft lottery since World War II. The Selective Service System placed 366 capsules, one for each possible birthday including February 29, into a large glass container. The capsules were drawn one at a time, and the order determined which young men would be called to serve in Vietnam. It was supposed to be the fairest possible method: pure chance, no human bias, no favoritism. Within weeks, statisticians noticed a problem...

In 2008, a security researcher named Luciano Bello noticed something strange about Debian Linux's implementation of OpenSSL. Two years earlier, a Debian maintainer had removed two lines of code that appeared to be using uninitialized memory, a common source of bugs. The fix looked reasonable. Static analysis tools had flagged the lines...

This is Part 1 of a 7-part series revisiting the trolley problem with deeper philosophical tools. Sometime between 1265 and 1274, Thomas Aquinas was working through a problem that had nothing to do with trolleys. He was writing about self-defense: is it permissible to kill an attacker to save your own life? His answer, in the Summa Theologica, introduced a line of reasoning that would quietly shape moral philosophy for centuries to come. The reasoning became known as the doctrine of double effect. Aquinas himself didn't formulate it as a tidy set of rules; later thinkers in the Catholic moral tradition, particularly Joseph Mangan in 1949, distilled it into the four conditions most philosophers now recognize. An action that produces both good and bad consequences may be morally permissible if: the action itself is not intrinsically wrong; the agent intends the good effect, not the bad; the bad effect is not the means to achieving the good; and there is proportionate reason for allowing the bad effect to occur. This sounds abstract until you realize it's the hidden architecture of the trolley problem itself. The Engine Behind the Thought Experiment When Philippa Foot introduced the trolley problem in 1967, she was, in significant part, probing the doctrine of double effect. The scenario was designed to test whether the distinction between intended and merely foreseen consequences holds moral weight. Consider the classic case. You divert the trolley to save five people, foreseeing that one person on the side track will die...

In 300 BCE, the Greek philosopher Epicurus had a problem. His teacher Democritus had argued that the universe was nothing but atoms moving through empty space, colliding and combining according to fixed laws. It was elegant, mechanical, and completely deterministic. It also left no room for free will, novelty, or anything genuinely new happening in the universe. So Epicurus introduced a small, radical idea: sometimes, for no reason at all, an atom swerves...

We started this week in a rabbi's workshop in Prague, watching the Maharal shape clay into something that could walk, work, and protect a community. We end it in a world where billions of people interact with golems every day, mostly without thinking of them that way. The golem tradition is roughly two thousand years old. The technology it describes, creating powerful servants from raw material and animating them with language, is roughly two years into its most dramatic acceleration. The gap between the tradition's wisdom and our current practice is where the most important questions live. What the Week Revealed Each post in this series mapped a different face of the golem pattern in modern technology. The spec-vs-intent problem showed that the golem's literal obedience is the oldest bug in existence...

When the Maharal's golem flooded the house, the Maharal didn't blame the golem. He couldn't. The golem had no capacity for blame, no understanding of what it had done wrong, no experience of having done anything at all. It followed instructions...

The Maharal had one golem. He shaped it from clay, inscribed the word on its forehead, and directed it personally. One creator, one creation, one chain of command. The relationship was simple enough that a single person could manage it. We don't live in that world anymore...