Humans need to call the shots in AI

AI helps us get things done faster, but it is not a miracle drug.

Over the past 24 hours, I’ve been working on an AI Art piece which required putting different people of various demographics together for a key visual.

The perennial issue with AI is that it cannot understand what you truly want, and you have to feed it suitable prompts and guidelines. However, this project was a true challenge because AI Art engines often struggle with more than one generated character.

After wrestling with Midjourney and Stable Diffusion for hours, I realized I was going about it the wrong way (ie. the way everyone uses AI Art tools to achieve finished artworks).

I had to do this the hard, human way – I designed the layout I wanted, then I generated each character separately, combined them together in my desired arrangement, and generated a suitable background for the montage. This required a combination of the AI Art tools, standard photo-editing and classical visual design principles.

This experience reinforces what I tell everyone – AI needs to be used alongside existing tools to get the job done. It can give you quick results, but they will be either generic (leading to mediocre outcomes) or will be not useful.

If you want a great outcome, you have to take charge and call the shots. AI helped me to achieve this visual in a day, and it would have taken me a far longer time to do it the old-fashioned way. But it still took a lot of time, thinking, debating and repeated iterations, and that’s what humans do best!

I help others grow through the power of words, visuals and technology. I also partner Oo Gin Lee to help businesses tell great stories.