This led to the first AI-related project I undertook that really blew my mind. I wrote a python script that read the captions, then sent them to ChatGPT with a prompt that asked it to get the exact coordinates of any locations mentioned, and to get dates for captions that mentioned a holiday or other even, but not a date, e.g.: "Christmas 1987," my birthday, which I included in the prompt, etc. I was then able to apply the results to the exif data the majority of the photos.

Now when I open Google Photos, my timeline stretches back to the day I was born, and the location heatmap shows the where photos were taken on extensive road tripping we did as a family.

But I generally know the dates far better than any AI could guess (based on ages of the individuals I know).

At least the scan date is a real piece of data that could be useful. This seems like it would cause more harm than good by polluting the data with nonsense, unless the added dates are clearly labeled as estimates

I guess this is assuming some scanning rig (camera scanning?) where it only takes a few seconds per photo. With a flatbed scanner, scanning the backs is almost certainly going to take longer than just typing dates in manually.

Technical background.

It's great to see stuff like this hitting the market!