Low Resource Computing 2026
摘要
文章介绍 2026 年 Low Resource Computing 研讨会（8 月 16-19 日，Dartmouth College 举办），回顾计算史：用 92kIPS 微处理器建国家银行系统，用 PDP-8 和 DAC 建渲染引擎，用 2MHz CPU、计算尺与梦想登月；指出软件浪费硬件的现象，如用 Electron 加载整个 Chrome V8 引擎编辑文本、用 gigawatt 数据中心问天气；讨论无限增长竞赛的不可持续性，强调 LRC 聚焦 kilobytes、kilohertz、nanowatts 规模的计算，主张将处理能力放回个人手中，构建更合理可持续的未来。
荐读理由
会议把过去微处理器和低资源成就作为启发，提出在资源有限时仍能完成计算，适合在资源约束项目中利用历史方法来减少对大模型的依赖。
原文
Dartmouth College Hanover, NH
August 16-19, 2026
Looking Back, Looking Forward
Computation started as a puzzle. We built national banking systems using 92kIPS microprocessors. We built rendering engines using a PDP-8 and a DAC. We flew to the moon with a 2MHz CPU, a slide rule, and a dream. The resources behind these accomplishments look modest only in hindsight. What is trivial now was enough then.
Somewhere along the way, we lost sight of this goal. Software wastes hardware because it can afford to do so. We use Electron, wrangling the whole Chrome V8 engine, just to edit text; we ask gigawatt data centers what the weather will be tomorrow. It's a competition to see how much compute and money we can throw at our problems before they fix themselves. Exascale datacenters keep growing, gigawatt compute farms keep chugging more electricity, and billion-dollar networks keep asking for more money. It's a race to the bottom, and other than being boring and expensive, it's impossible to keep growing like this forever.
This year, LRC is looking at the past as an inspiration for the future. As mainstream computers hog more and more resources, meaningful computation remains essential at the scale of kilobytes, kilohertz, and nanowatts. This workshop is all about countering the fad of infinite growth, it's about putting processing power back in individuals' hands, and it's about contributing to a more reasonable and sustainable future. How far can today's modest resources really go? Let's find out.
这条对你有帮助吗？