the diminishment of process

on copyright, LLMs and why there is no shortcut to good creative work

In July this year I went to Italy with my best friend. This has become an annual tradition: a handful of nights, near the end of the school term (around her work commitments) where I get her away from her young kids in a European city neither of us has been to. We eat and drink and walk around and we talk. Last year it was Bordeaux, where I ate a meal that made me cry and discovered dunes blanches. This time it was Bologna, which turns out to be perhaps the perfect city (not least because of its proximity to Modena).

My best mate is one of those terminally offline people; she has no social media, doesn’t live her life responding immediately to emails and might take several days to respond to a WhatsApp. I have never seen her show a YouTube video to anyone. If she phones me for a chat, it will almost always be when she is pushing a sleeping child around in a pram. I greatly envy the extent to which she is generally unplugged.

Our trip to Bologna was a stark lesson in what my over-reliance on my phone has done to me, compared to Elly, who in this instance operates as a sort of control group (of one—I’m not saying this experiment would pass peer review). Bologna is not a massive city; you can walk around the middle part in less than an hour. We were staying very centrally. She referred to Google Maps perhaps three times over a four-day period, and proceeded to construct a mental map in her head which she then added to via trial and error: ‘vaguely that way, I think’ / ‘I’m pretty sure we’re parallel’. It very quickly became apparent that I am completely incapable of doing this. To me, her ability to map the city in her mind was like some sort of witchcraft.

My sense of direction is quite famously bad. I get lost coming out of almost every public bathroom, unable to remember the route I took to get into it just moments before. I can’t orient myself in most locations and am terrified of getting lost; I don’t like just ‘wandering’ because I’m convinced we must be going miles in the wrong direction. My dad sometimes uses the names of roads that I have driven on dozens of times and I have absolutely no idea what or where he is referring to. I have to use Google Maps to get around the city I have lived in for six years now. And after four days of walking around Bologna I had next to no clue where my hotel was, or how we might get to anywhere we had already been. My orientation skills are at absolute zero.

It wasn’t always this way. I did the Duke of Edinburgh at school, and could functionally read a map. I went to Malaysia alone for three weeks back when you had to take a physical Rough Guide with you because your dumbphone would not be much use. I used to be able to drive from Rotherham to Surrey without the help of a GPS. I have no doubt in my mind that my navigation skills have dissolved because of my reliance on Google Maps, and the research seems to back this up: a 2024 meta-analysis of the available evidence, from the Journal of Environmental Psychology, says that:

GPS use is negatively associated with navigation ability, specifically environmental knowledge and sense of direction, indicating that the more individuals rely on GPS to reach destinations, the more poorly they perceive their navigation skills and the poorer is their knowledge of the environment.

Other studies show that it is specifically the hippocampus that is most effected by habitual use of GPS: Scientific American says that the hippocampus ‘is deeply associated with supporting spatial memory, spatial navigation and mental mapping’—as well as memory making, which perhaps explains why my memory is also so bad I can’t remember such things as having gone on holiday with certain friends only a handful of years ago.

Since that holiday, I have made a concerted effort to turn off my GPS and not use my phone’s maps when I’m walking or cycling, and when possible, when I’m driving. I’m trying to get used, again, to getting lost and figuring it out. I’m noticing, as well, more about my environment, and about the things I walk or cycle past every day; it’s making my days richer, in small but important ways. But I am still at a loss a lot of the time. It seems that when it comes to these things, you really do use it or lose it. And despite what a lot of people seem to want to believe, the same thing is true of many of the skills we’re outsourcing to technology at the moment. Not least of all, the ability to think.

---

A couple of weeks ago at the Edinburgh International Book Festival, I was invited to a panel on AI and creativity (which included the very brilliant Paula Westenberger). I was invited to be part of this even for a couple of reasons: one is that I am a writer of both fiction and nonfiction, and have been critical of the tech industry in the latter capacity (in this book and on this Substack as well). The second is that for two years I worked for the main writer’s trade union here in the UK, the Society of Authors, and during that period tackling the impact of AI on the creative industries was one of our main concerns, with copyright being a main focus. The central question of the session, as given to us, was this: Can AI ever be a force for good [in the creative industries] or is it destined to damage storytellers, creative practice and the art of writing?

Of course, there are myriad uses of AI that are a force for good, not least in healthcare; AI is a very broad term. There is a difference, though, between AI used for early disease detection and generative AI—and the latter is already contentious in creative fields. Most of you will know by now that at least two major companies have used enormous amounts of pirated material—books and research papers—in order to train their AI models (or, more accurately put, their large language models, which form a subset of generative AI). Specifically, the Library Genesis data set, which contains more than 7.5 million pirated books and 81 million pirated research papers, was used to train Meta’s Llama 3 LLM and an OpenAI model, though the latter claims that today’s ChatGPT did not use it and the staff members who used LibGen have since left the company (which makes it fine, then, I guess?). According to internal communications from Meta, these books are wildly valuable to these AI models which can’t be created without them, but the developers also aren’t willing to pay to license all that work or wait to go through all the paperwork. So they chose, instead, to access and use LibGen, which consists entirely of pirated work, which violates copyright law, and they did so in order to make money. The writers of this original material didn’t give consent and weren’t paid, and then everyone in the world was encouraged to use these LLMs to do things they could do themselves five minutes ago but apparently can’t do for themselves now—like writing their own essays / novels / stories.