The rise of generative AI needs to be put into context, so I'm going to explain the pros and cons of using AI in art creation. But first, let's reflect… I was born in the 1980s and witnessed the transition from computers as a novelty, to them being in most homes. As a child, I loved to draw and paint. I loved to explore everything you could do with computers and creative work.
As I grew up and started doing more serious digital artwork, I would often hear the “it's just pushing a button” comment when I people discussed digital art. I could easily brush that off and explain patiently, that a computer and Photoshop were like an advanced pencil. Just tools where I still did the creative work.
Today, I can’t say the same with such high confidence. Today paintings, videos, 3D models and music can all be made with just that… “a push of a button”. So the discourse becomes how can AI and human creativity work?
it’s an insult to human creativity to think that all of our artistic sensibility could be channelled through a single text prompt
It’s easy to come to the conclusion, then, that there’s no longer any need for artists, creatives and storytellers. But to me, it’s an insult to human creativity to think that all of our artistic sensibility could be channelled through a single text prompt. Art is intent, meaning, sensitivity and most importantly of all, it’s a reflection of the complexity of a human being. Being able to create a mostly random, beautiful image with the 'push of a button' doesn’t change that.
We’re still nowhere near a point in time where generative AI based on text alone can offer the same sense of control of a traditional workflow. For some projects, that loss of control is not a problem, for others that fine-grained control over any aspect of the process, is what makes or breaks the work. As OLIVER's Rod Sobral argued in his piece on AI in branding, human creativity with AI can succeed, but AI alone fails.
Outside of the debate about training data, which requires an article series of its own, what are the core pro’s and cons of generative AI? Let’s start with some cons and end on a more positive note with the pros.
Generative AI: the cons
01. Results can be frustrating
One of the main issues with generative AI systems today is that they’re great at making images that look good, but quite bad at making the image you had in mind.
02. Subscriptions are expensive
A lot of AI services are good at some aspects of image or video generation, and not great at others. If you want a more full-fledged experience, subscription fees add up unless you have the technical courage to install local AI models.
03. The training data is full of bias
There have been plenty of examples of how the bias is quite strong and most AI image generative systems. Several companies are trying to make more fair models, with varying success. In many ways it’s a reflection of the bias that already exists on the internet and in society in general.
04. The results are inconsistent
Everything constantly shifts around, and with most tools it’s impossible to do a “hold that thought” on elements that work, while continuing to iterate on other elements.
Generative AI: the pros
01. It’s great for brainstorming sessions
It’s very easy to quickly generate content for moodboards and to brainstorm on ideas. If you supply the ai systems with just a little bit of content outside of a pure text prompt, you can get some truly surprising and inspiring results.
02. There are a lot of tools and workflows
Chances are there’s an AI tool that fits well for exactly what you’re making. Especially the open source community around Stable Diffusion based systems like Comfy-UI and AnimateDiff are increasing the amount of creative control we have over Gen AI, on a daily basis.
03. Enables exploration of 'weird' ideas
By streamlining the art-pipeline in productions 'weirder' ideas can be explored, which would otherwise be too risky for budgetary reasons. AI tools could make productions less rigid. Need to change a main character's hair or relight a sequence? That would be very costly before, but maybe not as much tomorrow.
04. AI can be used as an amplifier
Tools like Magnific AI (as I explain in my article 'Don't fear AI') can take a relatively detailed input and add a final layer of finish to it, effectively working as a sort of 'amplifier' of your artistic intent. This is where I think generative AI is most interesting; much more so than in workflows that revolve around text prompting alone.
If you want to explore AI, read Martin's views on why generative AI is too generic as well as our look at Nvidia's approach to AI for video games. Looking for new hardware, read our guide to the new AI Copilot laptops, and why they may, or may not, be right for you.
