> [in 2020/2021] the dominance of autoregression was not as well-established as it is today: GPT-3 had turned some heads, but the ‘ChatGPT moment’ wouldn’t come until late 2022
I disagree with this. Decoders clearly worked. GPT2 was considered too dangerous to release, and I remember scrambling to get on the GPT3 waitlist. It worked.
But what made it go mainstream was RL. RLHF at first, then other improvements like DPO that were less of a pain in the ass to set up. Adding diffusion on top of that would be an even bigger pain in the ass.
So in short, diffusion never took off because it was just a more complicated way to generate tokens, and the real problem was getting tokens in the right distribution.
> [in 2020/2021] the dominance of autoregression was not as well-established as it is today: GPT-3 had turned some heads, but the ‘ChatGPT moment’ wouldn’t come until late 2022
I disagree with this. Decoders clearly worked. GPT2 was considered too dangerous to release, and I remember scrambling to get on the GPT3 waitlist. It worked.
But what made it go mainstream was RL. RLHF at first, then other improvements like DPO that were less of a pain in the ass to set up. Adding diffusion on top of that would be an even bigger pain in the ass.
So in short, diffusion never took off because it was just a more complicated way to generate tokens, and the real problem was getting tokens in the right distribution.
I wonder if we’ll get something like CDLMs for automated harness engineering, sort of piloting the LLM underneath.
how do tools like hermes do this? does it just review sessions and rewrite markdown files?
also haven't read too deep into the deepseek agent harness but the math in there was really cool. it sounded promising, at least.
"Attention is all you need" should be renamed into "Attention is sufficient but not necessary".
It’s the opposite: attention is necessary but not sufficient.
A quick look at the continuous diffusion models linked in the post shows lots of transformer models still