Was This Made with AI?

Or is this question oversimplifying things?

Artificial intelligence has become a powerful tool in many areas of human life, including creative and intellectual work.

AI systems are being used widely in medical and scientific research. Engineers use AI-assisted generative design to explore and evaluate possible solutions within constraints they themselves define.

The same principle can extend all the way into a small motorcycle workshop. A mechanic’s accumulated knowledge, including diagnostic procedures, unusual faults, compatible parts and years of practical judgment, can be organised so that an AI helps another person consult and apply it.

These uses are very different, but they have something important in common: their value depends upon what AI is being asked to do, the information it has been given and the judgment of the people using it.

That is easily forgotten because the most visible face of AI is often found elsewhere.

If most of your encounters with AI consist of scrolling through synthetic images, disposable text, imitated voices and advertisements promising to produce an entire book in minutes, becoming annoyed with it is understandable. Much of that material has been made quickly and accepted without much thought.

But this is only one use of AI, not its definition.

The difference becomes particularly important when we turn to art.

AI can be used lazily. Someone can enter a brief instruction, accept the first result and present it as finished work. It can become a shortcut around the very process from which worthwhile art normally emerges.

But artists are also incorporating AI into much more developed creative processes.

Refik Anadol uses machine learning and large collections of data to create immersive visual works. His installation Unsupervised was exhibited at the Museum of Modern Art in New York and subsequently entered MoMA’s collection.

Japanese author Rie Qudan openly used ChatGPT in parts of her novel Sympathy Tower Tokyo, which won the prestigious Akutagawa Prize.

Anna Ridler created and hand-labelled thousands of photographs of tulips for Myriad (Tulips). She then used that dataset to develop her AI-generated artwork Mosaic Virus. The labour involved in constructing the dataset was itself made visible as part of the work.

In music, Holly Herndon developed an AI system called Spawn, trained it with human voices and incorporated it into her album PROTO.

In none of these examples did the artist disappear and leave the machine to make the work.

The idea, intention and judgment still had to come from somewhere. Someone had to decide what the tool should do, recognise what was useful, reject what was not and continue working until the result expressed what they wanted it to express.

Whether we welcome this development or not, AI is unlikely to remain a separate technology that we can simply choose to avoid. It is increasingly built into phones, vehicles, agriculture, logistics, retail and countless systems operating behind the products and services we use every day. Much of this involvement remains invisible.

We can still question or reject particular uses of AI, and sometimes we should. But expecting to use only products in whose creation or distribution AI played no part is becoming increasingly unrealistic.

This does not mean that every use of AI is sensible or desirable. It means only that “AI was involved” is no longer a sufficient judgment by itself.

Against this wider background, this is how I have used AI in developing How We Are Here.

I have used ChatGPT as an assistant to help me organise my experiences, observations, concepts and philosophy, but also to examine ideas, challenge inconsistencies and search for clearer ways of expressing what I meant.

This did not happen through one instruction.

Over months of working together, I gradually shaped the way AI assists me in this particular work. I provided context, corrected misunderstandings, rejected unsuitable language and established the principles within which we work. The assistant became increasingly familiar with the voice of the book, its philosophy and the boundaries it must not cross.

Every word, every sentence and every paragraph in How We Are Here was subjected to careful scrutiny. Nothing was accepted simply because it sounded good. Meaning mattered, but so did rhythm, pacing, tone and the space surrounding an idea. Every word was placed deliberately to create the particular effect of meditative writing.

I cannot open an unfamiliar AI chat and expect the same results. It may produce an answer immediately, but that answer will probably be generic. It may introduce spiritual beliefs, metaphysical explanations or fashionable language that does not belong in How We Are Here.

The immediate answer is the easy part. Making the assistance useful takes sustained human involvement.

The images on this website were developed through a similar process.

I spent hours establishing a detailed visual protocol: the atmosphere, composition, perspective, lighting and colours; the position of the horizon; the amount of open space; and the way an image should behave as it slowly moves across the meditation screen.

Many images were rejected. Others were revised repeatedly. The instructions became increasingly precise until AI was able to produce images with the particular atmosphere and quality I wanted.

Although many visual themes are based on photographs that I took myself, AI extended what I was capable of creating. But it could only do so once I had learned how to direct it.

I am now applying the same principle to the development of a standalone AI tool for existential questions.

It would be easy to tell any AI to answer questions according to the philosophy of How We Are Here. It would also produce rather poor results.

The philosophy has to be translated into a working process. The AI needs clear instructions, distinctions, sequences, examples and boundaries. Its responses have to be tested, its mistakes recognised and its approach corrected repeatedly. None of that happens on its own.

Perhaps, then, a simple yes to the question “Was this made with AI?” tells us very little about the work or the human involvement behind it.

A more revealing question might be: What did a human being have to bring to the process before AI could become useful?

P.S. This article was written with the assistance of AI, pretty much in the role of an editor. The concept, ideas and sequence of this article came from me, while AI provided a platform for discussing them and helped me formulate them in clear and elegant English while maintaining my personal style. It took approximately two and a half hours before the article expressed what I wanted it to express. The entire process was paperless and used approximately 20% of my tablet’s battery charge.