ai-training agi-testing openai energy-consumption

Altman on AI energy: it also takes 20 years of eating food to train a human

SAM ALTMAN: “People talk about how much energy it takes to train an AI model … But it also takes a lot of energy to train a human. It takes like 20 years of life and all of the food you eat during that time before you get smart.”

old.reddit.com

February 22, 2026

1 min read

🔥🔥🔥🔥🔥

44/100

Summary

Sam Altman emphasizes that training an AI model requires significant energy, comparable to the 20 years and nutrition needed for human intelligence development. Demis Hassabis suggests testing AI by training it with a knowledge cutoff of 1911 to evaluate its ability to derive concepts like general relativity independently.

Key Takeaways

Sam Altman states that training an AI model requires significant energy, similar to the energy invested in training a human over 20 years.
Demis Hassabis suggests a test for AGI would involve training an AI with a knowledge cutoff of 1911 to see if it could independently derive general relativity.
Google has released Gemini 3.1 Pro, which includes new benchmarks for evaluation.

Read original article

View story →

Coverage & discussions

Hacker News · old.reddit.com

Altman on AI energy: it also takes 20 years of eating food to train a human ↗

Community discussions

Hacker News discussion ↗

Historical count: 5 · 5 saved in sample · Historical sample

Analysis ready

Community Sentiment

Negative

Concerns

Altman's comparison of human training energy to AI training energy oversimplifies complex issues, raising concerns about the implications for human value in a post-work society.
The significant difference in energy consumption between training humans and AI models highlights the flawed analogy, indicating a lack of understanding of the complexities involved.
Comments about post-human thinking reflect discomfort with the direction OpenAI is taking, suggesting a growing unease about the future of human roles in an AI-driven world.