Sunday Morning Coffee with AI. . .

 

 

Armed with a second cup of fresh, strong coffee, I messed around a bit this morning with artlist.io using its image to image function in an attempt to convert my hand-drawn map from September 2006 to something that more resembles an old map from the mid-18th century.  And just like my experiments with Ninja AI in June, the results are mixed.  

The above map is pretty good, but Artlist keeps fouling up the place names and has trouble putting a faint overlay of hexes across the entire area.  Hexes, admittedly, are not likely to be found on any genuine maps from the era in question, but there we are.  Frankly, I prefer the appearance of the Ninja map, but there were problems getting it to correct its errors.  Grrrr. 

As is the case with so much having to do with the various AI's out there now, the output generated is a direct result of the prompts entered.  For text alone, and when you develop a lengthy, highly detailed prompt, it is possible to get a reasonably good rough draft  to further revise yourself.  I've used Gemini, Claude, and Copilot to help with podcast scripts and things like lesson plans, refinement of student assignment prompts, grading rubrics for those same assignments, and the like.  But, when you spend that much time developing a two- or three-page prompt for the AI in question to (hopefully) get something useful back, you might as well just develop the document yourself.  Agreed?

For working with images, I am sure there must be a way to get the results I am after, but for now I find artificial intelligence to be kind of like a slightly dim child.  Eager to please, but hitting the mark only about 70% of the time.  Come to think of it, that rather stark assessment is probably true for most of us.  We get it right about two thirds of the time and skin our knees or drop the ball the rest of the time.  That's certainly the case for me.

Sadly, I have some real work to do now for an upcoming departmental fall kick-off in a little under three weeks on the 22nd.  Sigh.  So, further map experiments will have to wait until another time.

-- Stokes