By admin I’ve started experimenting with Suno and I have to say, I am very surprised! For a while, I was unsure of AI’s role in the art space, but the more I try out different LLMs, DALL-E, and Stable Diffusion, the more inspired I feel. There was a time when everything felt dead, as if we were headed into a new dark age of uninspired digital art and corporate logos. People who weren’t privy to an expensive, fancy art education or were unaware of the importance of art (thanks to a broken education system) have largely been gatekept from the field. I can imagine there are tons of imaginative and creative people driven insane due to a lack of expression.
As I continue to work on my game, I wonder how AI can be incorporated and accepted in this industry as well. I understand that AI itself is a morally gray area for the public at this point in time, but my belief is that this prejudice is mostly due to distrust and misunderstanding of the technology. It isn’t as easy as typing in a few words and hitting enter, as most people have been led to believe.
Prompting plays a large part in how any given model interprets and returns the data (visual, audio, text, or whatever data you’re working with). Being as specific as possible is extremely helpful, and limiting your use of negative prompts is also important. If you anticipate negative qualities within your finished piece by incorporating them into the negative prompt, the effect becomes something of a ‘ghost in the machine’. The AI now has an idea of the thing you wish to avoid, even when it wasn’t primed for that in the first place, so you end up with a sort of self-fulfilling prophecy. Negative prompting is most important in forcing away uncalled for or unintended qualities (poor hands, artifacting, etc.). The goal is to really nail the positive prompt so that the model doesn’t even need negative reinforcement.
The effect is similar in music AI prompting, as I’m learning. It’s very fun. I’ll keep this updated as I go!