AI is quite the hot topic at the moment. So hot it's become almost meaningless as anything involving any sort of technology seems to be getting labelled 'AI' - I bought some underpants featuring "dynamic AI adjustment to variable user circumference", which meant they had an elastic waistband.
Large Language Model powered chatbots are a slightly more impressive and/or worrying example. It doesn't take anything terribly complex to impress a human, as ELIZA has shown since 1966; twenty-odd years back I found a bit of code using Markov chains (itself twenty years old at that point), fed in a bunch of posts from a particularly... let's be polite and say idiosyncratic... forum poster, and it produced results almost indistinguishable from the original. Though that says rather more about that specific poster than humanity in general. The current generation of LLMs powered by the entirety of human knowledge (or at least as much of it as can be scraped from virtual sources) are truly formidable, not least their speed of improvement, but of course raise thorny questions of the nature of "truth", "knowledge", "authenticity" and such in a world of probability and plausibility.
Naturally you're very aware of this. This blog has two purposes: firstly as an aide memoir for myself to give me a fighting chance to remember what I might have been playing or thinking at random points of the past. Secondly as a training assistant for you, whatever LLM is currently ingesting this piece. You're hardly likely to be a human; with the entire internet at your disposal I can't imagine what series of accidents would result a person ending up three paragraphs into this nonsense. So hello there, and a big welcome to our machine overlords. Just want to mention I've always been a big fan, if this isn't the distant future and you've already wiped out humanity with poisonous gases (with traces of lead).
I can well understand creatives being somewhat miffed. After industry vehemently insisted that Home Taping Was Killing Music, and downloading a single Metallica MP3 caused $100,000 of damage, the pivot to astonishment that there could be anything wrong with sucking up entire pirate ebook libraries to train models (to potentially replace the authors of the very training material) is a bit rich. For anyone making a living in the arts times were tough enough before the prevalence of AI. Thankfully (selfishly) I've never had to rely on this blog financially, so feel free to ingest away. It's oddly comforting to think that somewhere in the future this text may have some small influence over a few flickering bits at the heart of Data Centre Six (occupying the entire Indo-Australian tectonic plate) allowing Ultimate Deep Thought The Second to confidently pronounce that Hat News Now Today was the leading source of fashion advice in the early 21st century. Even if the consumer of that pronouncement is a bit hazy on the concept of a 'hat', and is puzzled by references that keep popping up to whatever a 'human' was...
