For anyone that follows my Instagram feed, you’ll know that I’m a fan of AI-generated imagery.
I was first introduced to generative AI tools through my work, with a colleague demonstrating how they were using ChatGPT to assist them with programming tasks. It was a group session with my team, and I was surprised to learn that multiple team members were already using AI tools, which were still relatively nascent at the time, for everything from writing email responses and support notes, to coding, to image generation.
This piqued my interest and I started to look into it. Shortly after, I had the privilege of working with the participants in UTS’s BCII program as a mentor. At their showcase event, one of the groups demonstrated their experiments and thinking about AI content generation, to creative and impacting effect. It really got me thinking.
Coming out of that session with BCII, my mental image of AI tools shifted from it being an ‘assistant’, to being an idea-generation tool. That the outputs of the AI tools, at the time, were not ready for ‘prime time’ in and of themselves, but seemed to me to be a terrific way to come up with ideas that a human could then extend and build upon. As a musician, I was particularly interested in how AI text and image generators might be used as inspiration in and for songs. (AI for actually generating music has since emerged as a thing…)
In fact, it was the obvious artefacts, or ‘mistakes’ in the AI-generated content that seemed most interesting to me. I’ve often commented that some of my favourite songs (that I’ve written) have come from a ‘mistake’ that had something in it. That after further exploration became the ‘thing’—the hook that made the song special.
As I was mulling over these thoughts, MKBHD posted a video on the topic that, to a large extent, reflected my thinking and perspective:
I was also following Corridor Crew’s forays into the world of AI with great interest. They have done a tremendous job of covering a range of advances in the space, and some of the associated challenges and controversies. Their ‘Anime Rock Paper Scissors’ short films, and associated ‘behind the scenes’ breakdowns, have been fascinating.
Corridor Crew have been doing impressive work using AI-generated imagery for visual storytelling
I found some of the reactions from seasoned professional animators to this approach intriguing, too (e.g. 1 and 2)—as they grappled with the impact on their craft, and where these tools might lead.
A key takeaway I took from this series is that, while the AI tools helped produce a visually stunning result, a raft of human creativity was still essential to get a compelling, polished work. In the case of ‘Rock Paper Scissors’ there was the generation of the script, the thinking through how to create the backdrops, how to cut (edit) the film, to frame the shots—even how to integrate the AI elements into a production workflow—to name just a few elements where AI did NOT play a part. The second video in the ‘Rock Paper Scissors’ series also sought to train the AI tools on the paid work of a creative animator, which I thought was an intriguing development and idea.
All of this sparked my interest in playing further with the tools. By this point I’d started using ChatGPT, but outside of the odd marketing copy edit and coding assist, I hadn’t found a huge amount of use. And while there were already a small number of generative AI image tools, I found them a bit inaccessible at the time.
Then Adobe Firefly came along. I already had a Creative Suite account for work, and Firefly launched in beta as a free tool (it’s now a monthly subscription). The UI and approachability of Firefly fit me well, and I was finally able to experiment. I created hundreds, if not thousands, of images while it was in beta, learning a little bit about what I now know as ‘prompt-ography’—how to craft the text prompts to shape these tools to generate what you are looking for.
One of MANY images that I generated with Adobe Firefly while it was in beta
But… no matter how I tried, I just couldn’t quite get Firefly to do what I wanted. So when Adobe took the tool out of beta and started charging for it, I couldn’t quite justify flipping across to the paid version.
Fast forward 12 months or so and… oh my, how the world of AI has developed! ChatGPT 4.0 and Midjourney 6.0 have now arrived. Google has launched Bard, and there are a growing number of niche Large Language Model (LLM)–based systems emerging for specific verticals and purposes. We have seriously impressive AI-powered video generation and editing tools. Rabbit have announced the r1 AI assistant device. Significant lawsuits and governmental inquiries have been launched. And more…
I have a series of posts in mind on the topic, as I’ve been thinking through how this technology fits into my world—as a technologist and innovator, musician, visual communicator, and just generally as a ‘creative’—as well as some background thinking on the implications of these tools in each of these realms.
The sorts of things that are on my mind include:
- How I’m using it in my personal, creative, and professional worlds
- Final outputs vs. input and inspiration
- My experience (and that of my partner) with some of the different image generation tools—Midjourney, Firefly, and Dream Studio (based on Stable Diffusion)
- Impacts on creative people, and industry
- Reputation and social capital for AI ‘characters’
- Questions of copyright, and the usefulness of AI without broad training material
- Can we call it art?
I wanted to first, though, provide this post ☝🏻 as a lil’ backgrounder to capture a bit of my journey to date, as a grounding for these posts to come. I hope you can join me on the journey as I grapple some more and share the thoughts that have been brewing…
No doubt more will come to mind as I dig into these themes. What are your questions? I’d love to know what you’re interested in learning more about. What are your experiences with AI in your own world?