Failed and Bailed by Technology (with expert human assistance)

A Republic of Letters (emails now, actually) correspondent who had kindly read early drafts of my manuscript on uncertainty had suggested adding illustrations, perhaps using AI. I thought it was an unusual suggestion from a serious Swedish economist (although mainstream, he has written a fine book on Knightian uncertainty), but it resonated. Why not lighten up a geeky tract with whimsical images? I had something like the Little Prince or Curious George illustrations in mind. And as it happens, I cannot draw worth a damn, so using an LLM-interfaced AI that would follow my written instructions seemed like a dream. Possibly, for the first time in well over a year of trying, I might get something useful out of LLMs.

I already had several PowerPoint figures that I thought were “good enough.” What I thought I’d try to generate were images to introduce the four parts of my book, placing the relevant images on the first page of each part. 

Additionally, a visual theme emerged through conversations with a Republic of Zoom friend from Ireland. This friend is a broadly read intellectual, entrepreneur, and avid sailor. He suggested a nautical theme (which, given the subject matter, namely uncertainty) made a great deal of sense.    

Unfortunately, I simply could not get the several programs I tried to follow my image prompts. Yet rather than add the failure to my long catalog of LLM disappointments, I tried to see if the tech folks at Tufts who had been experimenting with the technology could help.

My inquiries led to an uber-helpful educational technology specialist, Freedom Baird, working at Tufts Educational Technology Services. Another stroke of luck – she was also an accomplished artist, writer, and educationist who knew about the nuances of book illustration. She kindly volunteered to spend some time with me on Zoom to see what we could do jointly.

If it’s too good to be true, it probably is. Certainly, in this case. We just could not get the LLM (I will not mention the brands) to produce anything worthwhile. Freedom said the technology was probably more useful for generating possibilities than actual images. Point taken, but I already had the possibilities. I knew what I wanted.

So, back to the old-fashioned approach of trying to hire an illustrator. But I am just as bad at finding and hiring people as I am at drawing. My sailor friend thought he might know someone who knew someone. I also pinged my network for suggestions.

Meanwhile,  and on a whim, I decided to troll through the Smithsonian and Met’s collection of Open Access images. I searched using keywords like boat, ships, quay, dock, lighthouse, and Columbus. I found several that could do what I had in mind, though the imagery would now be classical rather than whimsical. I created a montage using the images in PowerPoint, often adding whimsical captions, which I am allowed to do under the OA rules.

Freedom, who among her other talents is a Photoshop whiz, turned the PowerPoint slides into camera-ready images. (Illustration for Part I below)

What’s the coda: on the one side, the failure of LLMs to get the job done for me adds to my skepticism about the over-the-top hype about an AI breakthrough. On the other hand, there was a ton of technology – incorporating a lot of AI, I’m guessing, which was critical in getting the job: digitization, which allowed the museum art to be uploaded, search functionality, which helped me find, what I wanted, PowerPoint, Photoshop, Zoom (for conversations)… And most of this technology, including the AI elements, has evolved over the decades through much trial and error.

And, of course, human intelligence and human discourse, which helped me figure out what I wanted.

Part  I.

Invitation to the Voyage

Direct capture


1. Adapted from the Smithsonian’s Open Access online collection. The image to the left (c.a. 1750) is described as “Frame for the Title or the Dedication of a Music Book, or a Program of or an Invitation to a Concert.” The image on the right (ca. 1883-1884) is John Henry Twachtman’s Woman on the Quay, Honfleur.

2. Charles Baudelaire’s poem L’Invitation au Voyage was published in Les Fleurs du Mal in 1857, a book denounced as une outrage aux bonnes mœurs — “an insult to good manners” or “morality.” (Gotrich, 2018).

3. As mentioned in the Preface, we sail from Knight’s dock, not on Knight’s ship. Hence, Quai d’Knight.

Part II.

Formidable Obstacles, Forgotten Beacons

Direct capture

Compiled from images downloaded from the Smithsonian’s Open Access on-line collection. The bottom left image is Henry Wolf’s North-Easter, 1908. The image to the top right is Miner Kilbourne Kellogg’s Lighthouse, Civitavecchia, 1843.

Part III.

The Specialization of Entrepreneurial Initiatives

Direct capture

Compiled from the Smithsonian and the New York Metropolitan Museum Open Access on-line collections. Starting from the top: Kerr Eby’s, White Water, 1929 (Smithsonian); Types of Vessels series (N139) Racing Canoe, 1889, (Metropolitan); Nathaniel Currier, Pilot Boat William J. Romer, 1846, (Metropolitan); Nathaniel Currier, U.S. Frigate Constitution, n.d., (Smithsonian); Types of Vessels series (N139) Full Rigged Ship, 1889, (Metropolitan).

Part IV.

Imaginative Discourse

Direct capture

Henry Wolf, 1897. Open Access image from the Smithsonian on-line collection.