The Party Mountain [1]
A novella
I’m moving to Hornstull in October. It’s a very old house, built in the 1750s — from the street it looks like the Shire. It has a garden, and is close to the sea and the “Party Mountain”. Well, it’s just a mountain (by Finnish standards, otherwise a small rocky hill) but my landlords said they soon realised the Södermalm people go party there.
Perhonen lasikuvussa
Chapter 1
You probably know by now that AI is capable of writing stories. If you give it some input, like the start of a story, it can take over from there, and go on and on with generating more text, flawless in grammar and coherent in content, following a style so similar to yours that it’s uncanny. The storyline sometimes makes sense too. Put it this way, it looks like something you would have written, if you continued with the story.
Well, Dan told me about this text generation thing. Last night very late I mentioned my moving and the Party Mountain, he said it sounded like an award-winning title already. “When I get up I’ll run it through the Playground and see what happens” was the last message. It turns out Dan’s latest obsession is a new text generator that just came out, and his research company got access to the prototype. It’s capable of continuing with stories like never before, some fragments are indistinguishable from what could have been produced by human writers.
Dan also joked that if language describes the world, and text data embeds real-world knowledge, then linguistic plausibility could entail real-world plausibility — and the machine has taken in such a vast, vast amount of connections, encoded in texts — perhaps it knows something we don’t. Perhaps it could reveal ideas that are both novel and meaningful. You write down the past and it generates the future.
That would be interesting. I want to know how my life’s story unfurls, I want to know what the future holds, what’s in store for me at the new place and in what every now and then feels like another new chapter of life — even just plausible possibilities. In any case, it would be a lot more fun than horoscopes.
Ding ding… ding ding … ding ding
The sounds went off, who knows, a dozen times at varying intervals. These message notifications were my alarm. I feel kind of grumpy, or maybe just very dazed, but it’s not like that kind of dreamy state when I use to snooze over and over again, nine minutes apart, and have a dream (sometimes even a lucid one) between alarms. The messages are a succession of different versions of generated text. I skim through them, some parts are gold.
It’s 8:30am.
Quite a few user interviews in the morning, by lunch I’ve just had meetings with a design consultancy while using my neighbour’s WiFi and sitting on a cheap air mattress from the camping store.
Now looking through the generated texts in detail, AI has just told us the real name of the Party Mountain: Skinnarviksberget, “which is just above the Södermalmstorg square. It’s a big hill with stairs that lead to the top. It’s quite steep, but it’s a good workout for your legs. It’s a nice place to sit and enjoy the view of Stockholm from above.”
Meanwhile, Dan has sent through a few small pages’ worth of text generated from a section of Chomsky’s lecture on the “Cognitive Revolution”. The continuation is incredibly convincing in both style and content. Back then when I listened to recorded lectures of Chomsky, there was often that feeling of noticing the patterns of how he makes an argument. That feeling is a mixture of mentally nodding-along “yeah, makes sense” and a subtle wariness that the arguments are internally presuming and circular, and a sense of intrigue and curiosity in deciphering what is being said, even though the points are not immediately obvious. Well, now when I read the AI-generated text, I have pretty much exactly the same mixture of feelings. I’m intrigued by the text, but rationally I know why the points are not obvious, because, you guessed it, it’s mostly good at convincingly intelligible rambling.
That says something about how people talk, especially when they have to “perform” talking. But… It’s the listener who performs the perceiving and interpreting, and comes up with a feeling of “oh well, that makes sense”. It’s about the consumption of speech and text. What does it mean for content to have substance? Is it self-hypnosis of the reader who reads into it and between the lines? Or the listener who hears the sound of one hand clapping?
I don’t know… What does that say about us in general, when we are speaking? About thought? The AI-Chomsky tells of a story, a story about the automaton. It’s a kind of a thought experiment. Let’s suppose that I’m a robot, I’m in a room, and I’ve got a book in front of me, a rule book. The rule book tells me what words to put together to make sentences, and I’ve got a dictionary in the room, and I can look up words, so I look up the words in the dictionary, and I put the words together according to the rule book, and I can do that indefinitely, then what will happen is that I’ll produce sentences, paragraphs, and even a book that looks just like a book written by a human being, and it will be a book written in French, say, and it will be indistinguishable from a book written by a human being who speaks French.
“The question is, do I understand what I’m doing? Do I understand French? Does the machine understand French? And the answer is no, because I’m just following the rule book, I’m not really understanding French, just as an automaton following cascading sequences of routines would not understand French.”
Chomsky mocks the Turing test. Is appearing to be indistinguishable from an intelligent being, under some guises, a good test for intelligence? Descartes was dissecting large animals to understand physiology — scientists at the time even built machines that simulated digestion, but the goal was to understand digestion, they weren’t trying to establish that the machine was capable of digesting by trying to fool bystanders into mistaking the machine for a cow. I’m capable of digesting and I also think I’m capable of thinking. But how do I know what is really happening when I think? I guess other people can also think because they look like they can, just like I think I can.
The reason I came to Stockholm was to do the second year of my master’s in a “European Innovation” programme. Actually I met Dan because we were assigned to the same working group in an entrepreneurship simulation course. When we applied for the programme we had to do an IELTS test, to prove that we understand enough English. I speak Finnish, and Dan speaks Russian. Anyway, Dan once showed me a book which people use in Russian to prepare for the test. The test has a speaking component, and you have to talk to an interviewer about various possible topics and articulate your opinions. At the end the interviewer will judge whether you speak English well enough based on your ability to talk about things.
The test-prep book has a long list of topics, ranging from environmental issues to economic and social policies, and gives instructions on how to talk about each, with example answers for rehearsing. You don’t have to know anything about these topics, you don’t have to have an opinion on any of the issues, but you can convince the interviewer that you are able to talk about them.
When we talked about “resurrecting” dead philosophers by using AI to simulate their specialties, personalities and style (and at some point, have a convincing conversation with you), Dan brought up this sci-fi novel… it’s not well known at all in English. When I looked through the sparse, Google-translated reviews, I gathered that the novel was widely reviled by government censors (the novelist had gone into exile since, possibly in Berlin), critics, and common readers alike. Anyway, the plot is that in a sci-fi dystopia (possibly), scientists have managed to make clones of famous authors of Russian literature (reviews mentioned that the novel features sections imitating each of their styles). The clones — by the way, in my imagination there’s like a Terminator factory/warehouse scene where these clones are hibernating in tanks hooked up to cords and bathed in dim, cold industrial floor lights — are capable of producing texts in incredibly convincing resemblance to those of the literary giants. As the clones produce texts, a blue fatty liquid is secreted, which is for some reason very valuable, and it was sent to the past using a time machine… at that point I pretty much lost the track of the story.
In earlier times most people were illiterate. When the masses became literate suddenly there was this great thirst for things to read. Everything was read with fascination (partly, that we are fascinated by the magical phenomenon of transmitting thoughts through symbols) from newspapers to labels on laundry detergents. Then came my era, in which anyone can write, anyone can publish. There is an ocean, a galaxy, of text. In every minute more text was produced and put out there then you can read in a lifetime.
The company I work for pumps out a lot of “content marketing”, well, just by hiring freelance writers who can write about any and every possible subject to put together paragraphs after paragraphs, articles after articles to cover hot topics, about social trends, business, AI, or whatever. Dan’s company does it too. It generates leads — that is to say, the practice catches readers who are fooled into thinking we know about those topics. The texts of these articles are indistinguishable from what AI is now able to generate.
But is it about the reader who sees meaning between the lines? At every corner, you see the pathological relationship we have with the consumption of text, a race to the bottom of click-bait templates and algorithmic harvesting of impressions, of fabricated fascination. It’s a system that inevitably leans towards chaos and decay dictated by the second law of thermodynamics, as if we were all just particles in a lab, or the crowd in a stampede. An ocean of noise and indifference that drowns out anything that could be plausibly said. A colossal Plinko board that funnels each viewer-consumer to an individually constructed corner of reality. Everyone can publish but only publish to no-one else. Clearly it’s not the machines that created this algorithmic hell.
I’m trying to write my story about the Party Mountain. Does it matter if anyone reads it? Am I like a tree who rattles its full crown of leaves as the wind blows by, but it’s just all mechanics and not saying anything? Maybe one day AI can simulate for each of us an audience, even an interactive audience— that would make it full circle — not just spellcheck or speeding up “mental work of the manual kind”, or helping you find the right words, polish expressions, rearrange structure, and bring out the whole body of work. No more literary giants, no need for chasing achievements, for genius. Everything is a session of fun and games for your own entertainment.
People still play chess, and put years into studying strategies, building up knowledge and understanding, practicing — we try so hard to become good at things, don’t we? — when they know that even on a smart phone, there is a ready and patient companion player who can simulate playing at their levels, or any possible level for that matter. The machine can also analyse the whole history of every game and tell you what the name of the opening is, which move is good or bad, how things could have unravelled, and show you the possibilities. When you digest these instructions about patterns and potential future paths, you are “installing” in your memory and skill something that is aided by computational heuristics. Like putting on glasses or a prosthetic limb or incorporating a 3D-printed organ — you become in part made of a machine — only that this time it’s in software. What does it mean these days to play chess? What does it mean to get better or achieve the next level?
What does it mean to choose what to read? It’s not like a century and a half ago, when you could just move into the forest and read ancient epics in the original. Even a selection of books is more than to fill a life time, 60 plus years at a manic pace of one book per day. That’s only the volume up till now, in those 60 plus years who knows how many more thousand times will erupt into being. If you are what you consume, then you want the “good stuff”. To choose what content to consume, to choose what to do with life, to craft one’s own living and cultural intake and cultivation — that was once the idea of striving towards a good life. What does it mean to make worthwhile selections and choices? That is increasingly puzzling, and the increase is not incremental, it’s geometric, all-at-once, and overwhelming.
Meanwhile, we spend more time each day on materials that are particularly trivial, we seek sensations and novelty and entertainment, we don’t have the nerves to deal with having to craft or curate or select or choose, going with the tidal forces that wash us oblivious is effortless and comforting. Perhaps AI can help, let the resurrected Aristotle (or whosoever you like) give us a quote, a summary, a reference to consume just at the right time, in the flow of a conversation, of a “show”, when it’s mesmerising or entertaining.
What does it matter to distinguish between what is consumed and what is worth consuming? Between what is lived and what is crafted? Between the algorithmically constructed bubble or dream-like illusions and what is real and true? I don’t like where all this is going, but there is the mental life of what I like or don’t like, and then there is the phenomena of where the world is going, and I wouldn’t want to be too delusional either. Or? I don’t know…
