Abstract

Machine-learning-based AI is trained to mimic human research outputs. This challenges our traditional methods for assessing published research, because we have often relied on secondary signals (writing fluency) to set our degree of trust in the results. Fortunately, in mathematics it is becoming possible to mechanically check proofs, which largely meets this challenge. AI is also likely to increase the rate at which new research is created and published. How can human mathematicians assimilate all of these new results? In other fields where research has exploded, researchers have become increasingly specialized. This increases the need for scholars who can summarize results and communicate them across the specialized subfields. We will need more mathematicians to write review articles, monographs, and textbooks. Perhaps other forms of summarization and communication, such as knowledge bases, may become valuable. Can AI tools help with this process?