Simhash is an extremely fast and simple algorithm for detecting near duplicate text at scale which makes it particularly useful for deduplicating AI training datasets.