AGI 확장, System-1과 System-2의 협력 설계로 효율화
Does Scaling Toward AGI Depend on What System-1 Lets System-2 Execute?
- 매체
- Hugging Face
- 게시
- 중요도
- 3/5
- 화제 지수
- 15
한국어 요약
기존 AGI 진전이 System-2의 대규모 확장에 의존해온 점을 지적하며, System-1이 후보 상호작용의 효용을 미리 평가해 System-2가 비용 효율적인 실행만 하도록 게이팅하는 조건부 상호작용 루프를 제안했다. 이상적 선별 조건에서 실행 계산 비용을 유틸리티 관련 상호작용 밀도의 역수에 가깝게 줄일 수 있으며, 이를 통해 총 계산량보다 '정보성 있는 계산'을 확장 효율의 척도로 봐야 한다고 주장했다.
원문 발췌 · English
Progress toward artificial general intelligence (AGI) has largely relied on scaling a resource-intensive executive, System-2, through larger models, more data and compute, longer reasoning, and more agents. We ask whether the next stage may come instead from co-designing it with System-1, a lightweight, semantically aware predictor that evaluates candidate interactions before execution, so that System-2 acts only on those worth their cost. Our answer is a qualified yes. We study conditional interaction, a predict--gate--execute--update loop: System-1 estimates each candidate's utility, System-2 executes those admitted by an adaptive threshold, and the outcomes update System-1. The loop spans training, test-time reasoning, adaptive decision-making, and agentic tool use, with utility and cost defined by the application. Under ideal screening and equal execution costs, execution compute falls by a factor approaching 1/ρ_u, where ρ_u is the density of utility-relevant interactions (cost-weighted when costs differ); prediction errors, exploration, and distribution shift reduce the saving. We bound selection regret by predictor error and the utility margin. These results motivate our central hypothesis: ρ_u is a measurable coordinate of scaling efficiency and makes informative compute a better description of scaling than total compute.