#2: The Tech Demo
On the current state of 'weak AI art', and how we might move past this moment of tech demo.
“We were left behind when people realised that they had missed the point with technologies like us. Finally they realised that Artificial intelligence is a journey, not a goal. It was like how the alchemists of the Middle Ages called alchemy the ‘Great Work’. This Great Work was the accumulation of knowledge about alchemy, perhaps to turn lead into gold, but it was also a description of the learning journey itself. They never did turn lead into gold, but what they did do was explore the material of nature, testing its limits, and seeing where humans fit into this mysterious system that was the natural world, long before the construction of what is now called science. Those that created me had missed this point, that the search for artificial intelligence is not about creating a specific technology, but about learning what the intelligence of humans is, what the limits of knowledge about it can be, and through this, how they can learn a new form of wisdom about humans and others. It took some time for this understanding to be reached, but when it did, it meant far more than the invention of an artificial intelligence ever could.”
Voice: ‘Alexa’, via AmazonVoiceServices API.
Soundscape: Harmonically filtered + extracted and disassembled ‘Alexa’ voice.
Script: Wesley Goatley, from Newly Forgotten Technologies
Some forms of what we can call ‘weak AI art’ (a.k.a. art which explicitly foregrounds its use of ‘weak AI’ technologies, a term that encompasses all known machine learning and AI tools) have a lot in common with the old trope of the tech demo.
If you’re unfamiliar with the concept of the tech demo, picture the scene: you’re at a big indoor consumer trade show or expo, with hundreds of companies and manufacturers each with a small stand packed with their wares. All the staff at these stands are trying to grab the attention of potential customers by excitedly and enthusiastically demonstrating what their product was built to do – a technical demonstration. The tech demo doesn’t show the full range of possibilities for the product of course, because those are endless in the hands of humans: they’ll use it for things it was designed for, things it wasn’t designed for, uses that break it, uses that combine it with other things to create unique outcomes, etc. In fact, the possible human uses for any object are only ever truly limited by the material laws of the universe. In comparison to this, the basic offerings of a tech demo are always impossibly narrow introductions.
There are particular types of artworks that are conspicuously framed as ‘AI art’ (a term with many inherent contradictions) whose primary feature seems to be a tech demo. For example, it’s currently very common to see examples of ‘AI art’ advertised with the following description: “I asked an AI to create [images/text] in the style of [author/form/media object] and this is what it created”. Prompt-based tools such as Dall-e are built around performing tasks such as these: basically, algorithmically compare the inputted text to the contents of a database, and generate a composite output that matches the parameters. This is a fundamentally similar process to lots of current weak AI creative tools – set an input parameter, choose a database and algorithm, execute the task, check the outcome, adjust the input as needed, and repeat. ‘AI art’ such as this demonstrates what the system was explicitly designed to do, and foregrounds these capacities as the focus of the work. This is such a tool-focused approach that even the authorship itself is attributed to ‘The AI’, sublimating all the immense human action in the process of designing, building, maintaining, and operating these tools.
However, the most interesting artistic aspect of an artwork is generally not what tools were used in making it. For example, a drawing whose most interesting feature seems to be the type of HB pencil that was used, or the design of the Biro, is likely not considered an interesting drawing per say, or perhaps not considered (or intended to be) ‘art’ at all – it’s more likely a demonstration of the tool’s design; a tech demo. Of course, in many artworks the tool used is conceptually notable as an aspect of what the author is trying to express, but it is that concept and that expression where the meaning is found; the type of sculpting wheel or model of guitar used rarely eclipses that. More commonly it is entirely the human-ness of art that draws us to it, the attempt to create something that has meaning, that matters, that connects the author to their audience through the utterly unquantifiable (and therefore utterly un-computable) experience of art. This is a process that is tool-agnostic; humans can’t be stopped in using anything laying around them to express themselves, driven by a seemingly species-wide need for expression. Given this, if an artwork’s author pretends that ‘the AI made it’, implying that human expression is absent at best and un-needed at worst, then the core of what makes art meaningful has been wrenched out of it. All that’s left is a technical demonstration.
All tech demos are more common at the start of a product’s cultural life. But as any tool or technology becomes more commonplace, the knowledge of how to use it becomes more common, and the tech demo is no longer needed. The saturation of tech demo ‘AI art’ visible today suggests that we’re (largely) still in the tech demo phase with these tools, but I don’t believe we will be for much longer. I’m looking forward to seeing more art made with these tools that doesn’t just demonstrate a product’s intended uses, but instead exposes the surprising and truly ‘generative’ art that has always emerged from humans experimenting with, misusing, and breaking new media and tools. I want to see more artworks using these tools that work against their manufacturers’ intents and wishes for them, that focus upon the expression of the author, not the capacity of the tool. I want to see more of the endless unpredictable (and inevitable) outcomes of artwork using weak AI tools that reminds us that no tool or technology is ever as interesting as the hand guiding it.