I tried my hand at DALL-E-2, the AI image generator by OpenAI. When you input a phrase, the AI generates a selection of five images. I have chosen some of the better ones for this page. Take a look.
Phrase: An Astronaut in a Mumbai Local
As you can see, the Mumbai Local is a tad different from the Mumbai Suburban Railway we’re used to. It looks closer to the metro but isn’t the Mumbai metro either. The seats are different. But hey, the system is probably just learning the city. Here is another version.
Phrase: Sunflowers in Mars, Van Gogh style
Quite accurate. It looks like something van Gogh might paint; these are undoubtedly sunflowers and Mars is all red. I think I made it easy for DALL-E-2 to produce this.
Phrase: Tagore, Einstein and the Persistence of Memory by Salvador Dali
It probably did not get either Tagore, Einstein (maybe it’s a younger Einstein?) or the famous Dali painting right. See how “The Persistence of Memory” looks. As a standalone painting without context, I think one could take a long, hard look at it and feel some vague and undefined distress over the situation of humankind.
Phrase: Mahatma Gandhi Listening to Music, Vinyl
The man was an ascetic, but he did enjoy a bit of music, especially bhajans. As schoolkids, we were made to listen to his version of the Ram Dhun. He looks a a lot younger than we are used to, and more adorable than his real-life image would suggest.
Phrase: Indian Woman in a Space Suit, Wearing Fine Jewelry
DALL-E-2 generates five options for you to choose from. You can also modify the images a bit using the beta “Outpainting” tool which I didn’t find necessary. This image is beautiful, nevertheless. There is no reason why women in fancy jewelry can’t explore space. She ought to wear a helmet though.
Phrase: Man watching video leaning out of a Mumbai train
As a commentator on LinkedIn said, DALL-E-2 is likely to be confused: who on earth would watch a video while leaning out of a train window or door? Looks like it needs to learn the “spirit” of Mumbai a lot more. Two of the five options here, above and below.
Phrase: Indian Woman in the Countryside, in Crayons
It looks like something a child would draw, but it’s lovely. Another option below.
Phrase: Cats in a Cubicle, Van Gogh Style
It’s van Gogh alright. The cats are there, but cubicle, not so much. Maybe I should have said “laptop” or “computers.”
Phrase: Chinese Actors in an Indigo Flight Aisle
For some reason DALL-E-2 has chosen to represent a fashion show aboard a flight. No complaints, though. Was only testing the bot.
Phrase: Waiting in the Mumbai Rain with an Umbrella
It was raining outside my window when I ran this code, hence this phrase. I don’t know why the AI chops off the heads of the human figures in the background. Maybe to anonymize.
Phrase: Solitary House in a Snow Capped Mountain, Sunset
This has a photographic quality to it. If you want to use DALL-E-2 has an alternative to stock photography, you might want to use phrases that are easy to decipher and photographs that are widely available.
About the Experience
There is no doubt that DALL-E-2 helps you create some breathtaking images. You are perhaps limited by your own imagination, the availability of sample data, and some privacy and sensitivity restrictions. You may not find a lot of options for niche images or graphics, such as specific Indian archetypes. There are reasonable restrictions on usage of personally identifiable data or images. DALL-E2 is sensitive to the use of language and portrayal of racial, sexual and ethnic stereotypes, though it isn’t perfect all the time.
It’s good fun. There is a free tier which allows you to create 30 images in the first month and then 15 images every month thereafter. Try it. https://openai.com/dall-e-2/
Leave a Reply