Mercer: Training LLMs on AI-generated outputs triggers a race to the bottom
Chris Mercer · Backing Up to an Earlier Solution Changed Everything for Sean · Feb 13, 2024 · at 34:11
Measurement Marketing CEO Chris Mercer discusses the risks of model collapse and generic outputs when using generative AI indiscriminately.
0:00 / 0:07exact quote · 7.3s
▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark
“They've got AI, you know, language models that are being trained on the results of other AI language models, you know, and everything just kind of becomes a race to the bottom at a certain point.”
quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →
More from Chris Mercer
Mercer: Discard beliefs the moment they stop driving constructive action
“So is the belief useful? If it is, I'll keep it. And if it's not generating an action or a direction that I want to go, then I will really question that and say, well, just cause it was useful up until this point, it doesn't mean it still is. And I'm willing t…”