Hodge Conjecture

Mathematicians often study structures that can no longer be visualized. Such structures arise, for example, when one draws equations that link several quantities simultaneously. To understand them nonetheless, one describes their holes, edges, and surfaces using numbers and computational rules. The Hodge conjecture asks whether every one of these purely computational descriptions also corresponds to an actual figure within the structure. In other words: Does every matching calculation have a real shape behind it, or are there calculations without a geometric counterpart? The question has been open since 1950 and is one of the most famous unsolved problems in mathematics.

A million dollars for a proof

The Hodge conjecture is one of the seven Millennium Problems. In 2000, the Clay Mathematics Institute in the USA offered a million dollars for each of these problems. To this day, only one of them has been solved, namely the Poincaré conjecture. The Hodge conjecture is considered particularly difficult among experts, because no one knows in which direction a proof would even have to go.

However, it is not important because of the prize money. It is a bridge between two ways of thinking. On one side stands analysis, that is, calculating with continuous quantities. On the other side stands algebraic geometry, in which shapes are described by equations. A proof would show that both sides are looking at the same thing.

For practical purposes, this has no consequences for now. No program runs faster if the conjecture is true. Such foundational questions do have a long history, though: number theory was considered useless for centuries and is today the basis of all encryption on the internet. What would follow from a proof is only known afterwards.

From holes to equations

The starting point is a trick from topology. One counts the holes of a structure instead of measuring it precisely. A donut has one hole, a pretzel has three. These counts remain the same even if the object is bent or stretched. From them arise groups of numbers with which one can calculate.

William Hodge decomposed these groups of numbers into finer components. One can imagine this like a prism splitting white light into colors. Certain of these components are today called Hodge classes. Computationally, they have exactly the properties that a real surface within the structure would have.

The conjecture now states: every Hodge class can indeed be composed of such surfaces. Combinations with fractional numbers as factors are allowed here. This restriction sounds technical but is decisive. A stricter variant using whole numbers has in fact already been disproven. The conjecture has so far only been proven in special cases, for instance for structures of low dimension.

Between data center and lecture hall

You don’t encounter the Hodge conjecture in everyday life. It usually appears in the news when someone announces a proof. Such announcements come every few years and have so far never withstood scrutiny. An accepted proof would have to appear in a specialist journal and remain unchallenged for two years, as the rules of the prize require.

Currently, the term is of particular interest in the context of AI. Companies like Google DeepMind and OpenAI test their models on difficult mathematics problems. The Millennium Problems serve as a long-term goal and as a buzzword in press releases. So far, AI systems solve competition problems at school level or slightly above, not open research questions.

A more realistic role is that of a tool. Proof assistants like Lean check mathematical arguments step by step on the computer. Here, AI can suggest intermediate steps that a human then verifies. So if you read that an AI is working on the Hodge conjecture, this usually refers to this kind of collaboration, not a machine finding a proof on its own.