AI’s business model is starting to look like a snake eating its own tail: scrape human work, replace the humans who make it, flood the web with synthetic sludge, then wonder why there’s less quality data to train on. Silicon Valley calls it disruption. Everyone else might call it burning the furniture to heat the house.
I cant recall the source (I think it was a Telegraph article) but it only took current AIs 7 “rounds” to confuse its data amongst each other and it was 11 or 12 before it becomes completely corrupted. The testers basically used ChatGPT, Grok etc etc to search each other and they just had massive brain farts at the 7th round of questioning.
AI’s business model is starting to look like a snake eating its own tail: scrape human work, replace the humans who make it, flood the web with synthetic sludge, then wonder why there’s less quality data to train on. Silicon Valley calls it disruption. Everyone else might call it burning the furniture to heat the house.
Burning the furniture to heat the house is such a good analogy, wow.
Pretty sure some of that furniture was structural
Load-bearing, some might say
Shitting the nest is a more abrasive analogy for when you really mean it.
I think “Shitting where you eat” works a little better, given they’re shutting all over the internet, which is where their training data comes from
I’ve heard shitting on your own doorstep. Doesn’t fit as nicely as the furniture analogy though IMO 😆
I cant recall the source (I think it was a Telegraph article) but it only took current AIs 7 “rounds” to confuse its data amongst each other and it was 11 or 12 before it becomes completely corrupted. The testers basically used ChatGPT, Grok etc etc to search each other and they just had massive brain farts at the 7th round of questioning.
Best thing you could probably do at this point is make slop and pass it off as real work.