HackerPapers

arXiv:2104.03113cs.LG

Scaling Scaling Laws with Board Games

Andy L. Jones

Abstract

The largest experiments in machine learning now require resources far beyond the budget of all but a few institutions. Fortunately, it has recently been shown that the results of these huge experiments can often be extrapolated from the results of a sequence of far smaller, cheaper experiments. In this work, we show that not only can the extrapolation be done based on the size of the model, but on the size of the problem as well. By conducting a sequence of experiments using AlphaZero and Hex, we show that the performance achievable with a fixed amount of compute degrades predictably as the game gets larger and harder. Along with our main result, we further show that the test-time and train-time compute available to an agent can be traded off while maintaining performance.

arXiv comments
Journal reference
DOI
Subjects
cs.LGMachine Learning cs.MAMultiagent Systems

On Hacker News

Linked by 2 identifiable commenters in 3 comments across 3 threads, from Apr 2022 to Mar 2023.

Comments linking this paper, per year.

Discussed in comments3 threads

  1. Visual ChatGPT 1 comment

    • Model sizes are adjusted to reach a given quality. Many papers justified empirically the use of more neurons allocated to language than to visual understanding, like Imagen[0] (figure 4.a). as it yields better results. Debatably, it is also true in the human brain. To me, it…

      espadrineView on HNParent context

  2. John Carmack's new AGI company, Keen Technologies, has raised a $20M round 1 comment

    • > it isn't immediately clear that current RL algorithms will neatly scale with data & parameter count It may not be immediately clear, but it is nevertheless unfortunately clear from RL papers which provide adequate sample-size or compute ranges that RL appears to follow scaling…

      gwernView on HNParent context

  3. Tackling multiple tasks with a single visual language model 1 comment

    • > Scaling laws for RL agents are discovered, similar to language models. Still waiting for a followup to https://arxiv.org/abs/2104.03113 ...

      gwernView on HNParent context

Submitted to Hacker News

Connected through shared HN threads, not inferred similarity or a recommendation.