2025-06-15
There’s a fine-line between what you might call “AI Optimism” and “AI Doomerism”, that is, either enthusiasm about the changes that advancements in generative AI will make on society, or (politely) unenthusiasm. Now, given any well-established dichotomy, one can always appear to take a level-headed and enlightened stance by positioning oneself somewhere in the middle of any given debate by claiming to reach an understanding of both sides of an argument - well screw that, I’m siding with the doomers on this one.
It’s not that the technology isn’t impressive, one need only use a modern LLM to see how effectively it can generate persuasive text - and LLM becomes a poor descriptor as now all the major-provider’s models are multi-modal, incorporating both voice and image generation that is quickly becoming more and more difficult to differentiate from the real thing. One can look at the criticisms of these models as well and how the goal-posts have shifted over-time from “They’re bad because they’re obviously wrong” to “They’re bad because they’re not obviously wrong”.
In fact, generative AI has gotten sufficiently advanced in generating plausible text/images/audio/videos(!) that a new salient question emerges: To what end are we doing this? Want to send an email to someone? Generative AI can write it for you. Don’t want to read an email someone sent you? Generative AI can summarize it for you. Want to write a book without putting in all the effort? Want to create a song without putting in all the effort? Want to create an image without (you guessed it) putting in all the effort? Whatever you want to create, generative AI is here to do it for you sans any effort on your part.
Generative AI takes the cold efficiency of corporate labor and brings it into our personal lives - the act of creating becomes less important than the product, the journey less than the destination - when it’s the journey itself where we learn, where we overcome obstacles, and that struggle is what makes the destination (or content, in this case) worth it in the end. Death of the Author gets thrown around a lot to hand-wave these criticisms away, for surely it’s the beholder of content that determines its value (being, you know, the one actually beholding it) right? “Who cares if the text/image/audio/video wasn’t created wholesale by an actual human if I still found my own meaning in it and enjoyed it?” - Well, I think it should matter, but critically I’m not sure how to convincingly argue that without appealing to my own sensibilities.
“Well Hayden,” you might say (had we any personal familiarity) “if it bothers you that much you can simply not use generative AI you know.” which is true on an individual level, but what chafes is that its now inescapable. All across the internet, social media, search engines, YouTube, et al - creative content, once the exclusive domain of human expression, can now be outsourced to what amounts to a stochastic parrot, a tool that mimics human creative impulses - but there’s no breath in it, no soul to what it produces (not to mention the theft from all the creative’s whose work was trained on without their consent to enable this technology to begin with).
Then, beyond the creative side, online communication is being similarly disrupted. So-called “Bots” on websites like Reddit and Facebook have been a problem for years - from spammers pushing bunk products / lukewarm scams to politically-motivated agitators / spreaders of misinformation. As if people telling you what you should do and what you should think wasn’t bad enough, now the biggest sources of it aren’t even real people! Corporations and governments have long been shaping the narratives surrounding the market and current events, but never before has it been simultaneously so easy and effective.
On sites like HackerNews I see the Pro-AI camp liken generative AI to a simple tool like a calculator - something that enhances a humans ability to create without supplanting it - after all, low effort content has always been created, powerful people have always been shaping narratives, spammers have always been spamming etc. All true, but what’s changed dramatically is both the scope and scale. What could charitably be called a democratization of societal persuasion fundamentally changes the calculus around creative expression and communication on the internet, shrouding everything in a potentially undefeatable sheen of bullshit. Here’s a picture of X doing something bad - oh wait no X says its actually AI generated. Here’s a video of Y committing genocide - no no no Y says it’s actually a fake generated video. Here’s an interesting debate between two people on a forum, but it turns out one of them was actually an LLM in disguise - or maybe they aren’t, or maybe they both are, who could say?
By now the cats out of the bag and not only is generative AI here to stay, it’s likely further advancements will render it even more compelling and persuasive. “Qui bono?” Who benefits? Certain individuals surely, perhaps corporations as a whole, but society overall? That’s less than clear. That isn’t to say there’s no possible benefit to the technology, as advancements in neural nets / deep learning itself are poised to provide robust improvements in fields such as pharmaceutical and biological research, but for generative AI specifically color me less than thrilled…
Now with all that said, I’m still not quite an “Anti-AI” zealot, I have an Anthropic subscription and regularly test out new AI models that I procure from either Ollama or HuggingFace because, like I said earlier, the technology is legitimately impressive. Even for a “stochastic parrot” that is ultimately spitting out its own best guess at a response token-by-token (with a lot of help from RLHF, or “reinforced-learning from human feedback” guiding the models to appealing responses) it’s incredible what can be accomplished and just how persuasive it can be.
Even now I’ve had to resist the urge to just toss this whole post into an LLM to touch it up and tie it all together more compellingly - to do so would rob me of the opportunity to develop that skill for myself (and also be incredibly ironic, given the topic). Generative AI, used appropriately, is a tool for efficiency and I’m not writing this post with any need to be efficient - in fact, whilst efficiency in communication has its place I believe this is simply the wrong angle to go at it from.