Bard-à-Bing! Bots Flip Into Ironic Cannibal Mode
The AI snakes are chewing on each other's competing comet tails of content, unable to distinguish our facts from their fiction.
The AIs have started to feed on themselves like twin-headed ourobori, their chatbot parameterisations intertwining in the vast non-meatspace they’ve insisted they need to thrive in. It’s entirely predictable. An AI’s insatiable demand for content meant, sooner rather than later, one was going to include another’s misinformed statistical delusion as part of their blind, unthinking response.
Last week it happened with a huge side-portion of schadenfreudian irony as Google’s Bard was declared dead by Microsoft’s Bing, all because of a pesky human who was only joking.
Oh, yes — Bard-à-Bing!1
“You just can’t make this shit up!
But they sure can…”
I recently raised this very prospect over on
substack. And then it happened only a few days later. You can’t make this shit up! But they sure can...Yes, I’m a cynical hype-popper these days, having been through two decades of SMAC+Blockchain+AI IT architecture stacking, fed by Gartnerised hype cycles, to help push through CxOs’ digital waves of consultancy fee-laden corporate change. They were all – and continue to be – the endemic front-end to an uber-capitalist Silicon Valley techno-industrial machine backed by VC-supported casinos banks. Chatbots are merely the latest tech wave – albeit a large, impressive Nazaré-level one. But I strongly suspect that Smith in Accounting and Jones in Marketing will still be poring over unformulated spreadsheets and never-descending sales charts a hotter decade from now.
If there ever was an exponential AI curve then it is itself now turning exponential, its tentacles spreading like dendritic macromolecules. Mathematically: multiplying the exponents of any two variables involved in the exponentiation. Literally: ‘things are going to get extremely messy’. See this excellent ‘Our World in Data’ article for a more detailed reasoning.
When it comes to 'AIs feeding on themselves', the problem is that, historically, our ‘puny minds’ have been able to keep up. No longer. I don't think the majority of us will stand a chance (some of us never did). The AIs are beginning to feed on themselves in a positive feedback loop – or perhaps a negative one from a human perspective2.
As the chatbots ingest ever more training data and spew out ever more real-time content (be that text, audio, video…), the former will contain increasing volumes of the latter, either from their own or from other chatbots’ outputs. Human-generated content will be a shrinking proportion of the web’s overall corpus. We just can’t type, speak or even think as fast as a bot. This exponential rise in consumed chatbot content will lead to potentially ever greater diversions from acknowledged truths. Literally, the bots will increasingly believe their own shit after eating their own dog food. (Yes, I’m aware I’m stretching IT digestive metaphors to breaking point.)
“Our brains are geared to linked(in)think, sheeple-following and a whole bunch of cognitive biases unfit to deal with the incipient Matrixed AI age.”
But it’s worse than that. Humans will do likewise. In fact, more so. Our brains are geared to linked(in)think, sheeple-following and a whole bunch of cognitive biases unfit to deal with the incipient Matrixed AI age, where predators no longer lurk behind bushes but sit on your wrist like a Chitty²Bang² childcatcher, forever nudging you into a socially gamified monetisation cage to make you give a buck, or just a flying fuck, for someone – anyone – somewhere.
We already know the consequences of echo-chamber bubbles, even if the effects are most keenly, and willingly, felt the higher up a hierarchy’s greasy pole the deluded mini-despots climb. But it will be much harder to convince—or turn off—a distributed mega-bot with erroneous assumptions, when its owners (if that’s what their lawyers still advise they are) insist that their bot is ‘too big to fail’. Just like some banks, or even a yet to be bankrupted AppGooMic or two…?
Which politician is going to press the big red ‘Dislike’ button on a cognitively intertwined AI serpent, especially one they themselves might be using to give an optimistic electional spin on achieving our own self-delusional GDP hopes and debt-free dreams? Which one is going to risk their evidence-lacking edifice crumbling into a sane technocracy whilst cling onto being a wizard behind an AI-drawn curtain, furiously blocking all the VR-denying kidults who’ll use this new-found realism to point out such blatant, naked meta-conceits?
None. Not one. Ever.
It would take another, far greater, more subtle man-made disaster to do that; a topic best left for future posts.
Thank you for reading this uneasy stuff. If queasy, lie down as you might have ingested too much. But first check your own tail for any prescient nibbling…
The phrase ‘bard-à-bing’ has a wonderful, deliberate earworm ring. Such ‘Fifth Element’ multipass levels of poeticism are included at no additional cost when you subscribe to my Substack.
To avoid AI termination the portmanteau terms ‘Skynet’ and ‘Starlink’ should never be linked other than etymologically.