For this month’s 12 for ’26, Artificial Intelligence was the subject and I spent some time testing it with several different tasks.
Now it’s time for the retired teacher in me to grade AI on its successes and failures.
In Part 1, I shared a photo taken in Noble, Oklahoma perhaps around 1920. My father-in-law told me many Years ago that it was a picture of the first oil wells built in Noble. ChatGPT 4o did a great job telling me about the image and a bit of background about oil wells in Oklahoma. It couldn’t tell me when the first well was dug there. GRADE: B
In Part 2, both ChatGPT 4o and Perplexity were asked to identify the school building in Passaic, New Jersey from around 1920. Passaic had 12 numbered schools plus the high school and I even told AI that the school definitely was not #10 School. Neither AI came close to identifying the school as #12 School, Wilson Junior High. GRADE: F
In Part 3, I tried another photo taken c1918, this time one of a building that I believed was at Oklahoma University. Enough of the building was shown so that ChatGPT 4o made a correct identification of Evans Hall, the administration building (the third in the school’s history because the first two burned down). GRADE: A
In Part 4, the subject was a tintype photo of my 3X great grandmother, Mary (Scripture) Tarbox (1827-1866) and whether or not the picture was a post-mortem photo. Here’s where I think Ai can’t quite finesse details. While it addressed all the points I made, it said the photo was likely not taken post-mortem. I have other family photos taken in this era and not one of them is as strange looking as Mary’s. I still believe it was taken after she died. If she wasn’t dead, did the family have her photo taken to have something to remember her by because they knew she was dying? Either way, the photo is strange. GRADE: C
In Part 5, my last experiment, I shared two images from the Williams family in Dike, Hopkins, Texas in pre- and post-1900 photos. AI was asked to create a poem about John, his wife and small children in front of their new home and describe the daily life of the ladies in the second photo and to date that picture. I think AI hit the target with both photos. GRADE: A
So, how did AI do overall? Averaging the grades together, AI earned a solid B. It’s improving, but it’s strengths are in creating stories and identifying well-known public buildings. I was a bit surprised it couldn’t identify #12 School and still didn’t think Mary’s photo was post-mortem, but we agree to disagree there.
Moral of the story: I wouldn’t trust AI with its answers without doing the legwork myself to determine how accurate a response might be.
I’ve been following along in your series and agree that AI isn’t definitive and users must do their own legwork to assess accuracy. It’s just one tool in our tool box!