Skip to content

Instantly share code, notes, and snippets.

Show Gist options
  • Save charlesroper/64cf517fedd4bc2b060b514c4db19f63 to your computer and use it in GitHub Desktop.
Save charlesroper/64cf517fedd4bc2b060b514c4db19f63 to your computer and use it in GitHub Desktop.
Summary of Andy Masley’s AI articles.md

Summary of Andy Masley’s AI articles

A cheat sheet for why using ChatGPT is not bad for the environment

Here’s a detailed summary of Andy Masley’s post “Why using ChatGPT is not bad for the environment – a cheat sheet”, which distils his core arguments, data points, and rhetorical strategies:


🧭 Main Argument

Masley’s central claim is clear: using ChatGPT (or other LLM chatbots) has a negligible environmental impact, especially compared to meaningful climate concerns. Calls to limit individual chatbot use for climate reasons are a harmful distraction.


🧮 Personal Use: Individual Energy & Water Impact

🔌 Energy Use

  • 3 Wh per prompt is a conservative estimate (upper bound).
  • Equivalent to:
    • 3 minutes of a laptop
    • 30 seconds of a microwave
    • 1 minute of console gaming
  • $0.00051 per prompt (Washington DC rates) – almost imperceptible.
  • Masley compares this to buying a gum ball to illustrate absurdity: reducing ChatGPT use to cut emissions is like skipping a monthly penny purchase to save money.

💧Water Use

  • ChatGPT uses 10–25 mL of water per prompt (training + usage + cooling).
  • The average American uses 160 gallons per day (60 from electricity, 100 directly).
  • “50 ChatGPT prompts = 1 water bottle” sounds shocking until compared to 1200+ bottles used daily through other activities.

🧿 Relative Impact

Compared to Google:

  • ChatGPT is ~10× more energy intensive than a Google search.
  • But both round to zero in terms of climate impact.
  • Like comparing a digital clock (1W) to a watch (1µW) – the difference is vast in percentage, but trivial in practice.

Compared to Other Activities:

  • Skipping one shower second = enough water for 40 ChatGPT prompts.
  • One drying cycle = energy of 1000 goofy ChatGPT prompts.
  • The scale of these comparisons is key: ChatGPT's usage doesn’t register meaningfully on individual or collective energy footprints.

🌍 Global Use: Aggregate Misunderstandings

Data Centre Panic is Misplaced

  • Chatbots = only 1–3% of AI's total energy use.
  • Main energy hogs: recommender systems, predictive analytics, search and ads, computer vision, and voice AI.
  • ChatGPT’s total energy use ≈ 20,000 US households — small when compared with:
    • YouTube (2 million households),
    • Fortnite (400,000),
    • Microwaves used 10 seconds less per day globally = same energy saving.

Misleading Scaling

  • Critics say “If everyone used ChatGPT, emissions would rise!” — but this logic applies to anything.
  • Global use ≠ a valid reason to criticise personal use unless it’s a large slice of total emissions — ChatGPT isn’t.

🔧 Lifecycle and Production Costs

  • Training GPT-4 used ~50 GWh — seems large, but spread across 50B+ prompts, it’s ~1 Wh per prompt.
  • Embodied emissions of hardware (chips, data centres) ≈ 22% of lifecycle emissions — increasing per-prompt cost from 3 Wh to ~4 Wh.
  • But again, other devices (e.g., LED bulbs) have high embodied emissions rarely counted in similar debates.

🧠 Intuitive Analogies

  • Microwave metaphor: trimming a meal by 10s = 2 ChatGPT prompts.
  • Budget metaphor: spending time worrying about prompts is like budgeting by cutting 1p per month.
  • Dot visuals: one ChatGPT energy dot in a field of global energy dots = statistically invisible.

🚫 Counterarguments Addressed

  • “But it emits at all!” – Everything does. What matters is how much.
  • “But Google is better anyway” – That’s a separate debate.
  • “It’s still guesswork” – True, but so are the claims that it’s environmentally harmful. The best guesses say it’s minimal.
  • “This is whataboutism” – Not if the point is choosing effective climate actions.

💡 Strategic Climate Advice

Masley suggests we stop wasting activist energy on irrelevant targets and instead:

  • Fly less, eat less meat, buy green electricity, advocate for system-level change.
  • Focus on total emissions, not small relative increases in otherwise negligible areas.

🎭 Social and Cultural Observations

  • The disproportionate backlash against ChatGPT’s environmental footprint likely stems from a general cultural unease about AI, not the actual numbers.
  • He compares the scolding over ChatGPT to randomly attacking a particular phone game (“Wizard Clash 7”) for no reason other than cultural bias.

🧑‍🔬 Final Perspective

Masley is not anti-science. He presents himself as:

  • A physics graduate and former teacher,
  • Someone who’s deeply climate-conscious (vegan, doesn’t fly, walks to work),
  • Concerned with rigour, perspective, and effective climate action.

🧩 Bottom Line

Using ChatGPT is not bad for the environment. The numbers simply do not support the guilt. The climate movement should stay focused on what truly matters — not get sidetracked by symbolic but insignificant issues.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment