Open Thread

Adventures in AI:

I was in a Youtube discussion on the US Civil War, where a poster defending Grant’s management of the campaign claimed that the Union Army of the Potomac lost fewer men in the 51 days of the Overland Campaign than in 1861-1863. Not only is it a very misleading claim to compare just 51 days to 3 years, but on searching I found the Google’s AI answer varied and spat out contradictory information and information not found on the sources listed.

The first attempt (last weekend, when traveling, I can’t reproduce it now) gave a total of 647,000-some casualties, with breakdowns of 200-276,000 per year, with the KIA, WIA, and MIA totals listed. At first I thought “maybe this is including illnesses and accidents and deaths from those, but the funny thing was, that the total was sometimes FAR LESS than the sum of the KIA/WIA/MIA! When I went to the listed websites cited, I found nary a shred of this information. So where was Google’s AI getting it?

Now, every time I refresh, I get somewhat different numbers. Even then, most of the numbers are less than what I get by adding up the individual major and minor battles of a year. Say, it now lists 45,000-50,000 casualties for 1862 while by summing the individual battlefield losses yields a total closer to 77,000.

I fear for students using this crap to learn.