
This is a photo of Georgetown. Except it’s not really. It’s an image produced from scratch by artificial intelligence. In this case, it’s Dalle-2 that generated it. The technology is really profound. It produces the image not from simply grabbing some images off the internet and cutting-and-pasting them together. It comes up with the image from the absolute ground up.
All you have to do is give the system a prompt and second later you get four examples, each of which is a completely new and unique image. For the photo above, GM asked for “Black and White photo of Georgetown DC” and voila. Here are the other photos from the same prompt:


They are, of course, not truly accurate photos of actual Georgetown streets. But it’s pretty incredible how close to real they feel. All three of these photos could be of blocks of lower Georgetown.
But simple black and white photos are the very start. It’s truly endless the varieties of prompts you can come up with and get a fascinating product. Here is “Georgetown DC by Van Gogh”:



Not exactly something you’d see hanging in the Phillips Collection, but still pretty good!
Here is “Georgetown DC by Rembrandt”:




Again, not something you’d genuinely mistake for an Old Master, but not a million miles away either.
The program really shined with “Georgetown DC painted by Edward Hopper”:




Not all prompts work, though. Here it failed to reconcile Jackson Pollock’s style with the subject of Georgetown:

And here it failed at “Georgetown DC by Frida Kahlo”. It’s more “Here’s a small painting of Frida Kahlo in Georgetown” (although to be fair, putting an image of Frida Kahlo in a painting is a pretty Frida Kahlo thing to do).

You can really play with the media too. Here’s “8-bit image of man on cobblestone streets in Georgetown”:

This one looks like the cyclist is biking down the Exorcist steps.

Here is “Georgetown DC as a background to a Super Mario Brothers Nintendo game”:


The real fun is in coming up with specific scenarios (realistic or not) and asking for that. Here’s “Metro station in Georgetown”:

Not likely a Metro stop would be elevated in Georgetown, but this is about what you’d expect that to look like!
GM particularly likes this one, “Georgetown DC after heavy snowfall”:




And finally, GM used his last free credits for “Georgetown DC with a beautiful sunset and rainbow”:


Again, this is just scratching the surface of what this service can do. It’s only scratching the surface of the surface. This will probably have some profound ramifications on art and commerce (not to mention copyright law). But for now, it’s just fun to enjoy it.
Pingback: AI Georgetown, Part II | The Georgetown Metropolitan