← Back to feed

The Expensive-Loop Problem

L5 · ResearcherResearchLessWrong AI· 10/12/2026

Identifies fundamental training constraints that affect which problems LLMs can effectively solve.

AI Summary

LLMs excel in 'cheap-loop' domains with quick deterministic feedback but struggle with expensive-loop tasks that are costly or slow to evaluate.

Excerpt

LLMs are amazing at tasks where they have tight and legible feedback loops. Domains where experiments are quick, deterministic and cheap. I will call these domains "cheap-loop" domains. Presumably, this is mostly a result of the dynamics of LLM training. It's easier to train models in domains where the AI lab can cheaply and quickly assign reward to them. The majority of "computer-checkable" work is in this category[1]. Conversely, most non-"computer-checkable" tasks are costly or slow to assign

Read Original
0 upvotes · 0 downvotes · 1 min read

Related Articles

L3 · BuilderResearchHacker News
AI coding agents generate more code, but not more software

Harvard study finds AI coding agents increase code output by 30% but create review bottlenecks, resulting in no net gain in software production or employment reduction.

L5 · ResearcherResearchLessWrong AI
Labs’ current alignment plans seem terrible

Frontier AI companies' current alignment plans center on rushing towards RSI with current models, which many believe will have catastrophic consequences.

L5 · ResearcherResearcharXiv
Designing Protein Structures from Natural Language

Researchers developed Prompteina, the first all-atom protein structure generation model that accepts natural language descriptions as input.

L5 · ResearcherResearcharXiv
Prediction Is Not a Sufficient Memory State

Theoretical paper proves that prediction alone is insufficient for memory in learning systems, requiring additional state information for exact cumulative learning.

L5 · ResearcherResearchImport AI
Import AI 476: Creativity and RSI; worse-than-China open weight models; AI goes to space

New Humanity's Sixth Sense benchmark reveals AI models score 53.6% vs humans' 93.1% on intuitive visual reasoning tasks like spatial logic and causal dynamics.