I asked an A.I. to make a self portrait.

How does artificial intelligence see itself?

Mike Stevenson, MBA


A self portrait by Van Gogh hanging on a gallery wall.
Photo by Alina Grubnyak on Unsplash

Two things happened recently.

  1. In June, a (now former) Google Engineer, Blake Lemoine, publicly claimed an Artificial Intelligence program called LaMDA created by Google was sentient (Lemoine’s employment was terminated by Google in July).
  2. In July, DALL-E 2, an artificially intelligent visual design programme created by OpenAI that generates original images by request became available in public beta.

Regarding the first event, I won’t weigh in here on if Artificial Intelligence has achieved or can achieve sentience and/or qualify as a ‘person’.

There is loads of debate, speculation, and fiction that explores those questions.

Regarding the second event, I also won’t weigh in on whether an Artificial Intelligence can be creative — specifically, imagine appealling original ideas and realise them.

Instead, I wanted to mash these two concepts together and consider how would an Artificial Intelligence represent itself visually.

Basically, I decided to give an Artificial Intelligence program an eighth grade art assignment — make a self portrait.

Enter DALL-E mini

As the DALL-E 2 beta is currently invite only, I used an artificial intelligence web app called Craiyon.

Craiyon was previously called DALL-E mini, but changed its name at the request of OpenAI.

Craiyon notes some of its team also worked on OpenAI’s DALL-E program.

Craiyon is currently ad-based and free to use.

In sum, Craiyon was an easy option to choose for the task.

This is what happened when I asked the artificial intelligence program to create a visual representation of itself.