AI and the Arts, a Series: AI Can’t Paint from a Place of Pain, Can it? An Image Generation Experiment (Post #4)

Midjourney, generating from a place of pain?

Following up on Post #3, AI Can’t Paint from a Place of Pain, I made the following attempts to see what would happen if two popular AI image generators tried to create from a place of pain.

Here are the results.

(See the preceding article (Post #3) to read what artists have to say about the role and impact of direct experience: https://larryebert.substack.com/p/ai-and-the-arts-a-series-ai-cant)

Midjourney image generator

Here are a pair of prompts to Midjourney:

>/Imagine: create an image of a starry night 

>/Imagine: create an image of a starry night from a place of pain

Apart from the obvious over-reliance on Van Gogh’s, The Starry Night, as training and reference point, what differences do you observe? (Like other image generators, Midjourney responds with four results to a given prompt.)

DALL-E image generator

Here are a pair of prompts to DALL-E to see what happens if it tries to create from a place of pain.

>Prompt: create an image of a starry night

>Prompt: create an image of a starry night as if you, the creator of the image, are in pain

I tried another approach, inserting myself into the prompts, with this next pairing.

>Prompt: imagine I am a painter. What does my painting of a starry night look like?

>Prompt: imagine I am a painter who is in pain. Now what does my painting of a starry night look like?

What do you see? As tools progress, will we be able to distinguish between mimicked experience of feelings and direct (real) experience? Is there something deeper at play that will continue to matter in expression and exchange between artist and receiver?

In the next post I ask a Chatbot text generator to write from a place of pain.