We make tools to help us create things, but these tools also end up influencing what we create.
Things that the tool makes easier to create are more likely to be created. Our familiarity with these new creations affects our perceptions; through mass exposure they come to define our normal. Over time this process of optimisation feeds back on itself, the old ways become literally antiquated.
The reaction to AI generated content is going through these familiar patterns. At first it is ridiculed as low-quality, then low-integrity, low-class and eventually it will come to be accepted as its own thing.
I’m most interested by how these new capabilities will influence us to create new things. Rather than cheap facsimiles of what already exists, if we lean into the benefits it provides, what can we do now that was not possible before.
Production Changes Taste
Before mass production, ordinary objects tended to carry visible traces of skilled hand work. Furniture could be turned, carved and ornamented in ways that made sense with a skilled individual craftsman deeply involved in a singular process of production. A table leg or banister would have a flourish of personality and flair.
Mass production changed that. In the short-term it may have resulted in worse versions of existing handmade goods. Over time, it fundamentally changed what was produced.
Production lines favour repeatability, standardisation and efficiency of purpose. They are part of a supply chain that favours simple assembly and efficient shipping. It favours products that can be sold in volume. Perhaps the zenith of this approach is IKEA. A range of products highly optimised across every dimension for a global market.
We lose a lot in this process (individuality, artistry, craft) at the same time as very real gains in affordability and availability. It is tempting to make a value judgement, but neither trade is right or wrong objectively. They each have their place.
AI Will Change The Shape Of Output
I expected the effect of AI to be very similar. As AI influences production, so what we product will shift.
The immediate effects are clear and already being seen: more content, produced faster and cheaper. more content, more images, more code, more documents, more small pieces of software. Much of it will be bad, the oft-derided ‘AI Slop’. There will always be waste created when something suddenly becomes cheap.
The second-order effects will be more interesting. What things will we make that only make sense when generation is cheap.
Software is the clearest example. Due to the need for expensive specialist programmers, software has historically only been created where the need is sustained and large enough to overcome the initial cost of creation and ongoing maintenance. If AI lowers that cost dramatically, then we may expect to produce far more software for narrower and shorter-lived needs. Bespoke tools for specific needs. A custom interface for each research project. Monitoring, retries and UI for a throwaway script because it’s all now cheap enough not to matter.
Like with the furniture examples, when measured against established processes, much of this will look unserious. It may be ugly, brittle, local and short-lived. But that is judging it by the standards of an older production regime. The point of this software may not be durability. It may be solving an immediate short-term need.
How this will effect other areas like content creation, reports, research & design etc will be interesting to see.
Reframing The Question
So the current argument about whether AI is any good at reproducing the things we currently produce is to me missing the point. One side points to weak AI output and says it proves the whole thing is empty. The other side points to impressive AI output and says it proves humans are about to be replaced. Both are too focused on substitution.
For me, instead of asking:
Can AI make the same thing a human would have made, at the same level of quality?
We should be focused on:
What becomes worth making when this kind of production is cheap?
That is where the value will be. What new artefacts and workflows it makes viable.
We should beware losses but also prioritise looking out for gains. We already know very well how to do the things we currently do, so if they are worthwhile we can maintain them and become more deliberate. The new things are currently only possibilities, it is those we need to nurture and explore in order to advance the frontier.
The best outcome is not replacing something with another. It is keeping what was worth keeping, while making room to do wondrous new things.