2022

Compute Trends Across Three Eras of Machine Learning

Sevilla, Jaime, Heim, Lennart, Ho, Anson et al.

Understand

Compute, data, and algorithmic advances are the three fundamental factors that guide the progress of modern Machine Learning (ML).

  • In this paper we study trends in the most readily quantified factor - compute.
  • We show that before 2010 training compute grew in line with Moore's law, doubling roughly every 20 months.
  • Since the advent of Deep Learning in the early 2010s, the scaling of training compute has accelerated, doubling approximately every 6 months.

Built on

  • ( \bibnodate

    Klein, D · 2018

    Earlier work this paper cites.

  • ( \bibnodate

    Baidu Research · 2022

    Earlier work this paper cites.

  • ( \bibnodate

    GPT-Neo · 2022

    Earlier work this paper cites.

  • ( \bibnodate

    J, T., Sejnowski & Rosenberg, C.R · 2022

    Earlier work this paper cites.

Similar

  • ( \bibnodate

    Lieber, O., Sharir, O., Lenz, B. & Shoham, Y · 2022

    Cited alongside, same era.

  • ( \bibnodate

    nad Mikhail Pavlov, A.R., Goh, G. & Gray, S · 2022

    Cited alongside, same era.

  • ( \bibnodate

    Naver Corporation · 2022

    Cited alongside, same era.

Then

  • ( \bibnodate

    OpenAI, Akkaya, I., Andrychowicz, M., Chociej, M., Litwin, M., McGrew, B.Zhang, L · 2022

    Closest in time.

  • ( \bibnodate

    Rae, J., Irving, G. & Weidinger, L · 2022

    Closest in time.

  • ( \bibnodate

    Selfridge, O.G · 2022

    Closest in time.

Beyond the bibliography

alphaXiv searches the wider corpus for related work and actual follow-ups.

Open on alphaXiv

alphaXiv is searching for related work…