Dream-RSI: Recursive Self-Improvement through Evolving Worlds
1–10 of 59 posts
Re: Dream-RSI: Recursive Self-Improvement through Evolving Worlds
#2Would be very glad if anyone explained to me if and why this is revolutionary.
Re: Dream-RSI: Recursive Self-Improvement through Evolving Worlds
#3Would be very glad if anyone explained to me if and why this is revolutionary.
Here you go:
"Across algorithm engineering, mathematical optimization, and GPU kernel engineering, Dream-RSI achieves competitive or improved discovery quality while substantially reducing discovery cost in several settings."
Re: Dream-RSI: Recursive Self-Improvement through Evolving Worlds
#4Would be very glad if anyone explained to me if and why this is revolutionary.
Probably isn't by virtue of it being publicly released
Re: Dream-RSI: Recursive Self-Improvement through Evolving Worlds
#5Would be very glad if anyone explained to me if and why this is revolutionary.
Fairly certain all the labs are doing this (RSI) at this point. It's a question of how public their proclamations are about it and how they're positioning PR etc.
Even today's lighter weight models know how to write kernels and optimize them. I've had DeepSeek 4.1 Flash tune the crap out custom CUDA kernels on my own codebase and it was entirely competent at it. And cheap.
The innovation pieces will be in the harnesses to support this. Which I guess is partially what's going on here.
Re: Dream-RSI: Recursive Self-Improvement through Evolving Worlds
#6Unless I'm misunderstanding, calling this RSI seems misleading?
This looks like an optimization of current training methods, and a good one, but not "RSI" in the sense of a system that can perpetually improve itself forever.
Re: Dream-RSI: Recursive Self-Improvement through Evolving Worlds
#7Would be very glad if anyone explained to me if and why this is revolutionary.
Fairly certain all the labs are doing this (RSI) at this point. It's a question of how public their proclamations are about it and how they're positioning PR etc. Even today's lighter weight models know how to write kernels and optimize them. I've had DeepSeek 4.1 Flash tune the crap out custom CUDA kernels on my own codebase and it was entirely competent at it. And cheap. The innovation pieces will be in the harness…
It's not really RSI if you are just using the AI as a tool to help make it better. It has to be doing it itself, no? Otherwise self-hosted compilers are RSI.
Re: Dream-RSI: Recursive Self-Improvement through Evolving Worlds
#8Would be very glad if anyone explained to me if and why this is revolutionary.
Why would you assume it is revolutionary?
Re: Dream-RSI: Recursive Self-Improvement through Evolving Worlds
#9Earlier quoted context omitted.
Fairly certain all the labs are doing this (RSI) at this point. It's a question of how public their proclamations are about it and how they're positioning PR etc. Even today's lighter weight models know how to write kernels and optimize them. I've had DeepSeek 4.1 Flash tune the crap out custom CUDA kernels on my own codebase and it was entirely competent at it. And cheap. The innovation pieces will be in the harness…
It's not really RSI if you are just using the AI as a tool to help make it better. It has to be doing it itself, no? Otherwise self-hosted compilers are RSI.
That's my point about the harness.
Re: Dream-RSI: Recursive Self-Improvement through Evolving Worlds
#10This is a solid and very interesting paper! The authors were kind enough to publish the complete prompt for it (appendix B.1 on page 18), so anyone can try their approach with any LLM and see the results.