Papers & Research

OpenAI analysis: compute needed to train neural networks halves every 16 months

Original title:AI and efficiency

AI summary

OpenAI published an analysis showing that since 2012, the training compute required to reach equivalent performance on ImageNet classification has halved every 16 months. Reaching AlexNet-level performance now requires 44x less compute than in 2012, while Moore's Law over the same period would have delivered only an 11x cost improvement; the results indicate that for AI tasks with high recent investment, algorithmic progress yields greater gains than traditional hardware efficiency improvements.

Key Takeaways

  • OpenAI published an analysis showing that since 2012,the training compute required to reach…
  • Reaching AlexNet-level performance now requires 44x less compute than in 2012,while Moore's Law…
  • the results indicate that for AI tasks with high recent investment,algorithmic progress yields…

Quick Facts

Published
Updated
Source
OpenAI:官网动态(RSS · 排除企业/客户案例)
Category
Papers & Research
AI score
61
Read the original Read the full article at the original source
Why it matters

OpenAI 用 ImageNet 分类数据量化算法进步的速度,读者可据此比较算法效率与摩尔定律的贡献差异。

Related entities

Source & copyright

This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.

Original article · AIHOT entry

How to Cite

When citing this article, keep the title, date, and permalink, and credit Orange News as the source.

Orange News: OpenAI analysis: compute needed to train neural networks halves every 16 months. May 5, 2020. https://new.nat6.net/en/news/ctf4kaxoy1572sg8vh5ve9og6/

View all →