I have a confession to make. I didn’t write my previous post. ChatGPT did.
See, i couldn’t help but be intrigued by Jony Ive and Sam Altman’s announcement. Ive has been fairly critical of his own creations lately, going so far as calling them ‘legacy products’. We live in a world he designed, but as he told The Financial Times, ‘many of us would say we have an uneasy relationship with technology at the moment’. His disillusionment has been apparent for years:
When i first moved [in California], i came because it was characterised by people who genuinely saw that their purpose was in service to humanity, to inspire people and help people create. I don’t feel that way about this place right now.
‘Humanity deserves better’, on that we can agree, and he seems to genuinely believe that OpenAI holds the key to a better future. I’m less optimistic than he is, but if someone can bring good out of whatever the hell ChatGPT is, it might just be Ive. At least he seems aware of his tremendous responsibility as a designer: ‘if you make something new, if you innovate, there will be consequences unforeseen, and some will be wonderful and some will be harmful.’
That was enough to make me want to play with ChatGPT again. It so happens that i was stuck on a post about my experience making my first batch of notebooks. What began as an experiment to see if it could rework my draft ended as the publication of a brand-new article it wrote under my guidance. When i asked it to read my previous posts to adjust its tone and grammar, i learned that my ‘writing is witty and sardonic, with rhetorical questions, humour, and direct address — all combined into concise paragraphs full of parentheticals.’
It was a surreal experience, and not only because it pulled the anecdote about ‘eight-year-old me’ discovering his abilities out of thin air. (I did, however, borrow my Year 1 teacher’s Year 5 history lessons to read them at home. She was as pissed as she was impressed, and recommended i’d be transferred to a private school for ‘intellectually precocious’ children and that my ‘autistic tendencies’ were carefully monitored. The conflict between my parents on the ‘issue’ precipitated their divorce. So, you know, close.)
Trouble is, as i try to reread my post (its post? our post?), my eyes gloss over the page. It just fails to grab my attention. It vaguely sounds like me, if i was extremely tired, drunk, and writing about a topic i don’t know much about. Which is funny, considering that i haven’t had a drink in years and that it’s supposed to be an article about myself. The quote ‘good artists copy; great artists steal’ is often misattributed to Picasso, which i find endlessly funny1, and i’m not even sure ChatGPT is a good artist.
Writing using a machine is not the same as letting the machine write in your place, just like playing the piano is not the same as letting a player piano play for you. I’m not impressed by people telling me that the more creative you are in your prompts, the more creative ChatGPT will be in its answers. It reminds me of people telling you, with the vacant stare of the terminally stupid, that it’s not guns that kill people. Tools matter. Intentions matter.
The fact is that LLMs separate the material from the materials, the production from the practice, the creation from the creator. They squash the wonderful duality of the technè, the practical application of knowledge and knowledge of practical application that artists and artisans share. Almost nobody knows how LLMs work and nobody knows why they work. This is unprecedented.
The separation — i hesitate to use the word ‘severance’ — is what it’s all about or, more precisely, what it’s not about. The post isn’t bad, but it’s anodyne, which is way worse. So, yeah, without even having to talk about their social and ecological costs, it’s clear that LLMs aren’t for me. I’m still intrigued by Ive and Altman’s super-secret not-an-iPhone-killer device, though.
Humans have no universal faculty to judge aesthetics: Our appreciation of beauty is highly-contextual and depends on factors other than the raw visual stimulus. Most tech-workers are unaware of this fact, and for them, the fact that AI-art resembles human-art means it must be pretty damn good. But AI art is already in very poor taste: not just because it recycles existing conventions in a way that looks outmoded, but because it’s already overly associated with less-than-prestigious institutions. GenAI art has already reached polyester status, and this is just the beginning. Despite all the techno-utopian promises, our brains see it as ersatz.
Of course, GenAI is much worse for society than polyester. Synthetic fibers were bad for the environment, but so was the famously diabolical cotton industry. GenAI will first wreck the labor market for design professionals. But moreover, the tools are already being used to undermine the entire information structure of society in assisting the creation of disinformation that looks identical to reality. This, too, will damage its status value. Who wants to wear a T-shirt designed by the same software that powers the fake imagery used in authoritarian propaganda?
— W. David Marx, ‘GenAI is Our Polyester’, CULTURE: An Owner’s Manual, 25/06/03.
That being said, i still long for a better use of applied stats. LLMs are to the future of computing what CLIs are to the present of computing. They won’t disappear anytime soon as an underlying technology, but they need to disappear from the public consciousness. I don’t want to keep on rambling about the Knowledge Navigator, but we need graphical interfaces for artificial intelligence. Better yet, we need artificial intelligence for graphical interfaces, because i really shouldn’t be sorting my files in 2025 the same way i did in 1995.
We don’t need passive machines, which i what most recommendation algorithms are, soul-crushing and art-destroying devices that prevent you from ever discovering anything new. We don’t need reactive machines, which i what LLMs are, stochastic parrots that sing a narcissistic and sycophantic song. We need proactive machines, whose interfaces shape from you instead of trying to shape you, without ever forgetting the commonality of the human experience.
Bicycles come in different shapes and sizes because we have different needs and different bodies, but they’re always recognizable as what they are because they’re made for us. I wouldn’t know how to use the incredibly lightweight riding machine that Tadej Pogačar won three Tours with, but i’m pretty sure i’d be faster than him on my full-loaded cargo bike. If the computer still is the bicycle for the mind, let him have his race computer and give me my cargo computer. We’ll all be better for it.
- However, according to the painter Françoise Gilot, he often said ‘when there’s anything to steal, i steal’. The closest quote i could find comes from T.S. Eliot, who wrote in The Sacred Wood: Essays on Poetry and Criticism that ‘immature poets imitate; mature poets steal’. ↩︎