AI and Compute: Our analysis showing that the amount of compute used in the largest AI training runs has had a doubling period of 3.5 months since 2012 (net increase of 300,000x): https://blog.openai.com/ai-and-compute/
https://twitter.com/OpenAI/status/996782863797256192
AI分野、学習における計算
3.5ヵ月で2倍
2012年から見て300,000倍
指数関数的な変化

Moore's law - AI version
https://twitter.com/OpenAI/status/996782863797256192
https://twitter.com/IrwanBello/status/996806814548176896

Very cool analysis on compute vs year for some of the most notable advances in AI.
https://blog.openai.com/ai-and-compute/
https://i.imgur.com/HihjBx6.png
https://twitter.com/OriolVinyalsML/status/996792780927262720

We're in the super Moore's Law period: compute used on large ML projects has been doubling roughly every 3.5 months.
https://blog.openai.com/ai-and-compute/
https://i.imgur.com/tyD8qPZ.jpg
https://twitter.com/geoffreyirving/status/996784695814701056
3.5ヵ月ごとに倍増するスーパームーアの法則

2012年以降、AIの学習に使われた最大の計算リソース（学習中の総計算回数）は3.5ヶ月毎に倍のペースで増え続け、今後もしばらく続くと予想される。HWの改良、システムの改良、アルゴリズムの改善がこれを支えている。
https://blog.openai.com/ai-and-compute/
https://twitter.com/hillbig/status/996864638443708417