First, and most important, ML can identify outperforming equities based on patterns that would not have been selected for testing by humans. For example, ML can analyse all the responses of chief executives in earnings calls of the S&P 500 companies during the past decade. By going through millions of possible correlations, computers can identify patterns of good or bad investment performance for the company, or companies doing business in similar regions. Second, by enhancing natural language processing, ML can compare and contrast critical terms. That process allows data scientists to review, decipher and organise information in huge numbers of dense documents, like SEC filings or patent applications.

https://www.ft.com/content/efe4f97a-adb1-3cea-b098-2b616b5ce531?shareType=nongift