AI Building Exoskeletons - John Oestmann Music

Skip to content

AI Building Exoskeletons

One Rule to Keep Creative Inspiration

It has been suggested that technology’s evolution had two possible pathways from the mid-2000s: either Exoskeletons, or AI.

This is a huge oversimplification, but the point it tries to illustrate is this: Exoskeletons augment the actions that people already do. They allow people to walk over tough terrain, lift heavier things, be generally agile at a super-human level. Exoskeletons essentially represent tech, less as automation, and more as augmentation. It keeps the human at the center of decision-making.

On the other hand, AI here symbolically represents automation in general. It especially represents the offloading of cognitive work and decision-making. At its extreme, it is telling an AI to make a piece of art, then coming back later to receive the art. In this extreme, the human’s "hand" touches very little of the process. They are no longer involved.

I don’t think actually seeing AI / Exoskeletons only as these extremes helps anyone (both can be very positive), but when our north star is to make sure that Creative Inspiration is being served in the world, this can be a useful framing.

Creative Inspiration is a nebulous force. It is highly personal and can resonate inconsistently at different times. But it does seem that one of the commonalities behind the creation of Inspiring works relies both a sense of Intent and Craft of the Creator. This is one reason why AI-generated arts feel so offensive to so many people: it scrambles and mis-attributes the Craft and Intent to the person who releases the piece, when so much of both comes from, at best, a giant stew of training data, and at worst, stealing another artist’s style or work, uncredited.

I wonder if leveraging AI to build Exoskeletons helps us here.

I’ve been testing and looking into how I should bring Generative AI into my own workflows. I had been largely AI-Skeptical until a couple weekends ago, when I experienced the state of AI in the Software Programming space. I have seen how, even if AI development were to stop today, it is now at a level enabling iterative prototyping and adapting-tools-to-local-conditions that was previously unfeasible. We’ve began seeing legitimate Open models reaching the frontier, giving GenAI back to the people whose data it was trained on, and diffusing the power that was previously held by the duopoly. I have also started hearing reports about data centres implementing closed-water-recycling systems – a sign that there is very active work on making the infrastructure more environmentally friendly. Still a complex issue with many arms and legs, my scales have now tipped from leaning-negative to leaning-positive.

I didn’t want to release anything AI-generated until I had tested and thought through some of the ramifications first – especially to make sure I am not putting something out that hurts Creative Inspiration. I am writing this blog as I am currently working on my first deeply AI-assisted release: major updates to The Weave Explorer.

Through this process, I am finding myself settling on a middle-way: use AI to build Editors that act as Exoskeletons for ourselves to still be the artists of our work. Agentic Programming enables more of us to create bespoke software exactly for our workflow. It allows us to add features as we need them. By building Editors / Exoskeletons it maintains four boons:

We keep artistic intent and nuanced decision-making about what makes up the final art.

We draw from the Craft we have already developed, and we extend its reach.

Editing Software can be shared with others, open-source. If others love your workflow – you can give it to them. They can adapt it to their own environment.

Editors are deterministic, AIs are not. Editors allow us to repeat the same workflows with precision, not rolling on lucky-tokening.

There is a a very possible risk that this "AI-Generated Art vs AI-Generated Exoskeletons" framing will appear out of date within days, but I am partly writing it as helpful reminder to myself. This is not the first time I have reached a similar conclusion. There is more nuanced exploration and discussion to have around this (e.g. why aren’t we applying the same respect to software programming itself? Why has the software industry embraced GenAI far more than other industries?), but in aiming to set foundational guidelines to protect Creative Inspiration, I do think this is a solid framing to start with.

Related:

Location / Album:

Blog