588

LESSWRONG
LW

587
Language Models (LLMs)Recursive Self-ImprovementAI

1

ASI-ARCH: “Does this hold up?”

by DataDeLaurier
26th Jul 2025
2 min read
0

1

This post was rejected for the following reason(s):

  • No LLM generated, heavily assisted/co-written, or otherwise reliant work. LessWrong has recently been inundated with new users submitting work where much of the content is the output of LLM(s). This work by-and-large does not meet our standards, and is rejected. This includes dialogs with LLMs that claim to demonstrate various properties about them, posts introducing some new concept and terminology that explains how LLMs work, often centered around recursiveness, emergence, sentience, consciousness, etc. (these generally don't turn out to be as novel or interesting as they may seem).

    Our LLM-generated content policy can be viewed here.

  • Insufficient Quality for AI Content. There’ve been a lot of new users coming to LessWrong recently interested in AI. To keep the site’s quality high and ensure stuff posted is interesting to the site’s users, we’re currently only accepting posts that meet a pretty high bar. 

    If you want to try again, I recommend writing something short and to the point, focusing on your strongest argument, rather than a long, comprehensive essay. (This is fairly different from common academic norms.) We get lots of AI essays/papers every day and sadly most of them don't make very clear arguments, and we don't have time to review them all thoroughly. 

    We look for good reasoning, making a new and interesting point, bringing new evidence, and/or building upon prior discussion. If you were rejected for this reason, possibly a good thing to do is read more existing material. The AI Intro Material wiki-tag is a good place, for example. 

1

New Comment
Moderation Log
More from DataDeLaurier
View more
Curated and popular this week
0Comments
Language Models (LLMs)Recursive Self-ImprovementAI

A Breif Look Into An Automated Architecture

The paper AlphaGo Moment for Model Architecture Discovery introduces ASI-ARCH, a system that spits out 106 new linear-attention architectures that beat today’s best. The tone is pure hype—“superintelligence,” “AlphaGo Move 37,” the works. The engineering is real, though, so let’s peel off the marketing sticker and see what’s actually under the hood.

TL;DR  
ASI-ARCH is a very fast, very expensive grad-student simulator. It proposes, codes, debugs, and tests architectures at scale. That’s useful. It is *not* the birth of a new kind of mind, nor a universal law of discovery. Think “industrial assembly line,” not “Newton’s Principia.

What Works

1. Three specialists in a loop  

Researcher LLM: scans past runs and the literature, then suggests a tweak.

Engineer LLM: writes the training code. If it crashes, it reads the stack trace and fixes itself—no human in the loop. 

Analyst LLM: digests the metrics, compares siblings, and writes a short memo that the Researcher reads before the next round.  

2. Cheap screening, then expensive validation

20 M-param “toy” models run 1,773 experiments to skim the cream. Then 400 of the most promising are re-trained at larger scale. Twenty-thousand GPU-hours sounds insane, but staged filtering keeps the budget low(ish).

3. Learn from your own notes, not just the textbooks

Section 5.3 shows later models improve more from the system’s own internal notes (“analysis”) than from the distilled human papers (“cognition”). That’s a neat sign the loop is genuinely exploring, not just remixing lecture slides.

Where the Story (paper) Overreaches

1. LLM-as-judge is squishy 

The fitness score blends accuracy with an “architectural quality” grade handed out by—you guessed it—another LLM. That invites “persuasive writing” hacks: if I can *explain* the idea eloquently, the judge may inflate the score. The authors don’t show calibration curves or inter-rater agreement, so we’re flying blind.

2. Scaling law for science” is just a straight line on one graph

They plot GPU-hours vs. SOTA architectures and get a nice linear fit. That’s a local trend, not a universal law. Scientific progress is non-liniar; sometimes nothing ever happens, then a paradigm flips overnight. Calling a single search curve a “law” is like calling one hill a mountain range.

3. This isn’t Move 37

AlphaGo’s famous move was surprising, simple, and conceptually deep. ASI-ARCH’s winners—PathGateFusionNet and friends—look like Swiss-army-knife stacks of gating, routing, and residuals. Effective? Sure. Beautiful? Not really. It’s brute-force refinement, not lightning-bolt insight.

4. “ASI” is just branding

Artificial Superintelligence conjures images of a system that can do everything humans can, but better. ASI-ARCH is a narrow, compute-hungry architecture search tool. Slapping the ASI label on it muddies the water for everyone.

So What Should We Actually Care About?

Self-correcting code is finally practical

Watching an LLM debug its own CUDA is wild—and it scales. If you can define a search space and a score, you can throw GPUs at it and get new artifacts. 106 new SOTA architectures land in the community’s lap. Some will be duds; some will seed the next breakthrough. Either way, it’s a win for us researchers.

ASI-ARCH won’t wake up tomorrow and write a better ASI-ARCH. It will let a small team churn through more ideas than a conference hall full of post-docs. That’s not superintelligence, but it’s a very powerful tool in the right hands.