AI and efficiency

📰 OpenAI News

AI training compute requirements have decreased by a factor of 2 every 16 months since 2012

intermediate Published 5 May 2020
Action Steps
  1. Understand the trend of decreasing compute requirements for AI model training
  2. Analyze the implications of this trend on model training costs and efficiency
  3. Apply this knowledge to optimize AI model training processes and reduce computational costs
  4. Explore opportunities to leverage more efficient AI models in product development
Who Needs to Know This

Data scientists and AI engineers can benefit from this insight to optimize their model training processes and reduce computational costs, while product managers can leverage this trend to plan for more efficient AI-powered products

Key Insight

💡 Algorithmic progress drives more efficient AI model training, outpacing Moore's Law

Share This
💡 AI training compute requirements decrease by 2x every 16 months!

Key Takeaways

AI training compute requirements have decreased by a factor of 2 every 16 months since 2012

Full Article

We’re releasing an analysis showing that since 2012 the amount of compute needed to train a neural net to the same performance on ImageNet classification has been decreasing by a factor of 2 every 16 months. Compared to 2012, it now takes 44 times less compute to train a neural network to the level of AlexNet (by contrast, Moore’s Law would yield an 11x cost improvement over this period). Our results suggest that for AI tasks with high levels of recent investment, algorithmic progress has yielde
Read full article → ← Back to Reads