The Canada goose is a common sight around the city, specially when the weather warms up, and the migratory birds return from its overwintering grounds in the south. More than a dozen species have been identified. While they eat insects, they spend a lot of time eating grass and other plants. They enjoy hanging out in all kinds of environments including on top of buildings downtown, but they are often found in parks, beaches, marshes, and so on. During the months of May through August, they lay 2-8 eggs once or twice per breeding season.
I recently took this photograph of a Canada goose, who lay all by itself on a grassy lawn. I thought it looked like the kind of pose one may find in a biology book. This gave me the idea to add descriptive labels using AI and test the “intelligent” capabilities of the software.
I uploaded the image to the AI software (which I shall leave unnamed for now 😉) and asked it to “overlay information on the Canada goose using naturalist motifs and font from the 18th century.” I told the software that it “should contain the colors of the different parts of the goose, including a small translucent map showing the north American region where the Canada goose lives.” I did not specify the color or the parts of the goose that should be labeled. I included the instruction to “add a fancy bird-themed frame with naturalist flair around the picture.” The exact nature of the art wasn’t a concern in this instance because I was more interested in the accuracy of the technical information.
The results are outstanding. The labels are generally correct even though I did not explicitly include them in the prompt. The “fancy bird-themed frame” came out rather fancy with thematically appropriate imagery such as feathers, eggs, and marsh plants. I was delighted and thrilled by this level of reasoning. Upon closer inspection, I noticed that something wasn’t right with the image. If you look at the map, it shows the distribution of birds all year in the USA, not Canada. Doh!
I had a further idea to test the intelligent capabilities of the software. I gave it this prompt:
Keep this image but add a summer distribution in orange on the map encompassing Canada
Not only did the software add the orange colour as per my request, it also updated the map legend to include Summer in the corresponding colour, and the word 'Breeding' in parentheses, which I did not ask it to do, but it’s brilliant.
The Latin words and the jotted note at the bottom of the image are a bit much design-wise, but they are playful in a tongue-in-cheek kind of way that is perfectly suitable for my informal approach. These additional elements can easily be altered or removed by just asking the AI to do so.
One drawback was the deterioration of image quality in the second iteration with the orange distribution.
Overall, I am excited about the results because it shows a level of reasoning and processing that is fairly complex. I keep thinking about how long it would take the average person to create this image ‘manually’ using their knowledge of art, graphic design, biology, photography, map making, typography, history, and so on. That's not counting the years of training in college and practical experience at work that are required to reach this level of competence. It must be exciting being a teacher or student with this technology. Humanity is facing an explosion in intelligence, a Renaissance if you will, and it’s about honking time.
| X | InLeo | NFT Showroom |
