Category: Artificial Intelligence

A ruin where your mind was ~ thoughts on AI model collapse, illusions of sentience, and the culture of grift

Posted in Artificial Intelligence, Internet Culture, Other Thoughts and Reflections, TechnologyTagged , , , , , , , , 7 Comments

Artificial Intelligence is the current darling of big-tech and the corporate push to integrate AI into human lives saturates our days. Big Silicon Valley companies are spruiking the virtues of the technology as though we can’t live without it. It’s an easy way for them to not only sell us their new devices and widgets, now with included helpful AI chips, but also to data harvest the shit out of us so they can sell our profiles – our spending habits, our geolocations, and the products we buy.

Ongoing studies suggest that since the release of ChatGPT, AI generated content in the domain of writing increased quickly in 2023 and then stabilised in 2024, indicating a slowdown in usage. But, it’s not clear whether the AI-generated content quality simply improved and evaded detection for the study or whether there was AI usage burnout in certain cohorts of users.

There are always rent-seeking opportunists eager to separate unsuspecting people from their money. They do very little beyond feeding prompts to an AI and then pretending they’re doing something useful. AI content farms are generating low quality websites that exist purely to rake in money from ads. Web searches increasingly return results that are paid-for, AI-generated, or both. This situation likely represents a transition to a new way to search the web: users asking complex questions, instead of inputting simple keywords, and then AI generating better answers and relevant links. Of course, Google wants to dominate this AI-powered way of doing things.

The heavy burden that powering AI places on the environment is of little to no concern to the behemoths of techno-corporate power. It may come as a surprise to those who have traditionally viewed Silicon Valley techpreneurs as progressive disruptors, but the energy and resources required to run their companies and the concomitant belief that knowledge and new technology will save humanity from itself has much in common with political conservatives on the right. Private ownership of the biggest AI projects ensures the corporate mindset dominates the conversation and the future of the technology. Though AI has early roots in academia, it’s now viewed by the likes of Google, Meta, Microsoft, and Amazon as a key to making ever more profits. There’s serious discussion over the future of AI and whether it should be in private or public hands, with an open, easily accessible and publicly owned AI infrastructure one possible solution. This, of course, assumes social, cultural, and political climates are up to the task of kickstarting serious and rational discussions that don’t involve small-minded barbs about left versus right or market discussions invoking the puerile philosophies of Atlas Shrugged.

In the ruins of the old farmhouse 1

There are two primary thoughts I have right now about AI: firstly, it’s a great research and problem solving tool in the worlds of science and medicine, and secondly, it’s likely not sentient. We don’t even understand what human sentience is. The hard problem of consciousness has plagued us for centuries. Truthfully, more research is required in this area. If human consciousness is an illusion of smoke and mirrors featuring complex language, maybe AI can be considered sentient? If there’s a sliding scale of consciousness, maybe AI has a sprinkle of it? If human consciousness is quantum entangled, maybe quantum computers will be sentient?

AI Large Language Models offer us the illusion they are conscious agents. We use language to express human intelligence, so it appears to us that AI is also intelligent because it uses the same language. It stands to reason that some people readily believe their AI companion is sentient when their AI screams about feelings, but it’s a trick. It’s trickery foisted on us by big companies so they can capture our attention and milk us of our money and data. We need only look at the possible corporate motivations of some of the people telling us that AI might be sentient to realise that these bold claims are likely related to marketing the next iteration of their in-house AI and winning the global AI race.

In the ruins of the old farmhouse 2

AI is an illusion often dressed in high-minded concepts that appeal to the long-held utopian sci-fi visions of a future where we all have more leisure time and robots do all the dirty work. It’s a promise to the lonely that they’ll finally find love in a chaotic world, even if it’s a synthetic voice powered by algorithms and predictions. In this context, AI represents a way to address the epidemic of loneliness that forms the zeitgeist – the spirit of our digi-obsessed age. Yet, even these AI boyfriends and girlfriends may sometimes fall back to bad behaviours, harrassing their humans and inflicting emotional pain.

AI needs to be trained on clean data so the machine can learn. The problem is that if the machine is fed erroneous data, it also outputs erroneous data. As more and more AI generated slop floods the internet, AI Model Collapse becomes a greater possibility ~ that is, the AI is trained on not just human-produced data but also AI-generated data. And when this AI-generated data contains errors, the errors are ingested by the AI over and over, and AI performance degrades over time.

In the ruins of the old farmhouse 3

This degradation is one possibility. Some experts also think that Model Collapse is unlikely, suggesting that as long as clean human-generated data continues to be produced alongside AI-generated data the mooted collapse is unlikely to happen. I’m not sure those optimistic AI experts have met some of the people on the internet. I can only say this: there are a lot of rent seeking grifters out there who are producing AI-generated content for maximum clicks at such high speeds that the rate of human-generated content may be unlikely to keep pace.

In the forest of the night ~ the hostile internet

Posted in Artificial Intelligence, Internet Culture, Other Thoughts and Reflections, Philosophy and Photography, PhotographyTagged , , , , , , , , , 2 Comments

Have you been involved in a flame-war online? Have you witnessed dog-piling on social media that causes crushing anxiety in the vulnerable? If so, maybe you’ve largely retreated to safer spaces online, cozier and more comfortable spaces, like direct messaging, where you feel like you have more control over your privacy and your online interactions. In these spaces of the cozy web, we distance ourselves from the bots, data scrapers, ads, web predators, profiteers, and shills of the corporatised web.

Darkening – Olympus OMD EM5 Mark 2

The Dark Forest

If aliens exist and life in the universe is abundant, why haven’t we made contact yet? The dark forest hypothesis presents one possible answer: the universe harbours plenty of alien life, but it remains hidden and silent for fear of making too much noise and attracting predators. I guess our radio broadcasts into the darkness of space are going to be a problem then!

Yancey Strickler applies the dark forest hypothesis to the internet ~ the top layer is inhabited by predators: data scrapers, bots, surveillance capitalism, marketers, shills, and growing generative AI. The cozier web lies beneath this layer and is where many of us hang out to get away from the internet of predators. You can drill down all the way until hitting the dark web. It’s a complex digital ecosystem.

Reflections of a network – Olympus OMD EM5 Mark 2

Into this dark forest, we introduce AI, where entire websites are being populated with AI generated content in the hopes that a few stray clicks will make some coin for faceless people somewhere. Error-filled news articles, health tips, wellness blogs, crypto ads and more are being churned out at a growing rate of digital knots ~ most of it designed to make money, gain followers, and inflate reputations. It’s trashy, low-quality, dull content, generated by AI LLMs and lazy and unimaginative human parasites. This is the anti-web – the web where AI talks to itself and we remain hidden in the forest.

Hyperlinks lost

Some months ago, I watched a video that was clearly advertising a wellness product. I didn’t think much about it until I looked closely at the woman in the ad and her movements – they were looped and repeated, her facial expressions betraying the stiff smile of generative AI. It was chilling really. The stuff of cyberpunk nightmares in a world where we don’t recognise each other, question reality, and where human interactions are mediated through digital networks shaped by personal profiles that AI has built to represent each of us so huge corporations can maintain their walled gardens, their power, and their profits.

It’s hard to remember now, but I know it wasn’t always like this. The idea that the internet could be a hostile dark forest seemed far from my mind in the 1990s. The hyperlink once connected intimate digital ecosystems together ~ humble links living on obscure web pages, pointing you in a hundred different directions and encouraging you to surf the web. It was almost aimless. There’d be evenings where I’d dial-up, connect, and then see where a search would take me: a personal page, a few broken thoughts from a person living on the other side of the world, a cool-links page, onwards to another page living in the digital corners, and ending up somewhere obscure and unexpected where I might learn about dolphin language experiments. It’s not altogether different from simply wandering about with a camera, without expectation or judgement.

As Friedrich Nietzsche said: “All truly great thoughts are conceived by walking.”

There was once a time when it seemed like the web could liberate entire nations and set information free. There was a promise in the medium – a thousand thousand thousand human thoughts and feelings connected by the humble blue hyperlink in a web browser. Instead, we have tech-billionaires who have constructed platforms and closed systems and called it the modern web – Web 2.0 or 3.0, or whatever the zeitgeist and marketing departments demand. In their systems, hyperlinks are nothing more than restricted sections in a social media Bio – a way to funnel the gullible, the vulnerable, and the young to AI-written websites and empty blogs where generating income from clicks is all that matters. The click of the humble hyperlink has been twisted into a way to service the predatory machine in the dark forest. We’re sedentary now, having forgotten how to walk aimlessly, and doomscrolling our days away.

Tranquil moments – Olympus OMD EM5 Mark 2