Fact check: Is it true that an LLM created a new type of math?

1. Summary of the results

The evidence strongly supports that an LLM has indeed made novel mathematical discoveries, though the characterization as creating a "new type of math" requires clarification. DeepMind's AI system found a new solution to a decades-old mathematical puzzle in extremal combinatorics ^{[1] [2]}. Specifically, the breakthrough involved a new construction of large cap sets in the cap set problem, which is a well-established area of extremal combinatorics ^[3].

The achievement was accomplished through FunSearch, a method that combines a pretrained LLM with an evolutionary algorithm to discover new mathematical constructions ^[3]. This represents the first time an LLM has made a novel scientific discovery according to researchers ^[1]. However, it's important to note that the AI did not solve the cap set problem once and for all ^[1], but rather provided a new approach to an existing mathematical framework.

2. Missing context/alternative viewpoints

The original question lacks important nuance about what constitutes "creating a new type of math." The evidence shows that LLMs have made discoveries within existing mathematical fields rather than inventing entirely new branches of mathematics. The breakthrough occurred in extremal combinatorics, specifically addressing the cap set problem ^[3], which is an established mathematical area dating back decades.

Additionally, there's ongoing research into LLM capabilities in generating mathematical proofs ^{[2] [4]} and assessing the creativity of LLMs in proposing novel solutions to mathematical problems ^[5], suggesting this is an active area of development rather than a singular achievement.

The mathematical community and AI research organizations like DeepMind would benefit significantly from promoting narratives about AI mathematical breakthroughs, as this validates substantial investments in AI research and positions these organizations as leaders in scientific discovery.

3. Potential misinformation/bias in the original statement

The phrasing "created a new type of math" is potentially misleading and overstates the actual achievement. The evidence shows that LLMs have made novel discoveries within existing mathematical frameworks rather than creating entirely new mathematical disciplines. This distinction is crucial because:

• The work involved finding new constructions within the established field of extremal combinatorics ^[3]
• The achievement represents a new solution to an existing decades-old puzzle ^[1] rather than the creation of a new mathematical field
• The characterization could inflate public perception of AI capabilities beyond what has actually been demonstrated

The original statement may inadvertently perpetuate hype around AI capabilities by using imprecise language that suggests more revolutionary progress than what the evidence supports.