Michael:

As with all “learning” systems, AI depends on there being a source of good information from which to learn. This source may be termed a “teacher”, “editor”, or “verifier”. Today’s general LLMs are fed masses of information, primarily derived from scraping the Internet, without any qualifications or filtering. You and I cannot imagine teaching our children by exposing them to clearly erroneous, biased, and hate-filled information (as well as the true and valuable stuff), without any context or judgment about the material presented. That would be malpractice of the highest order! And yet, OpenAI and its competitors do exactly that, without compunction.

Historians (at least good ones) start with an educational background that equips (and compels) them to critically assess a document, and to contextualize it, before embedding it in a historical framework. They do this so that others may more easily recognize patterns, causes and consequences, and be informed about choices they may make in future circumstances. Historians also record events in particular ways that enable future historians to do all of the above (e.g., Heather Cox Richardson’s “Letters from an American”).

So, no, I do not think AI as presently constructed, is in any danger of replacing historians, or of doing what historians properly do. The issue, however, is that some non-historians think that Josef Goebbles was a historian, and mistake propaganda for history (a project of the Soviet Union, of Mao’s Cultural Revolution, and of Trump’s 1776 effort, as well as other authoritarians.)

Keep up the good work. It’s sorely needed.