AI Sustainability, The New Yorker Way

Treat our NYU AI systems like the busy New Yorker you are. Get to the point, be clear, and don’t waste anyone’s time. Each extra word you type needs to be analyzed and generated. That takes energy, and energy isn’t free. It’s like when you leave your cab idling just a little longer than necessary—those minutes add up, and so does the fuel bill and your meter.

It's estimated that a single conversation with ChatGPT (10-50 prompts) can "cost" around 500ml (16oz) of water for cooling the servers (Li et al., 2023). But while your prompt is “cheap”, the cost of unnecessary tokens and context accumulation is not tiny at scale. 

Every unnecessary word is a resource wasted.

We know that worrying about the environmental cost of a single prompt is a distraction—like trying to save the planet by ending your shower a second early. However, AI's resource consumption scales exponentially as it integrates into every application we use, guaranteeing that collective energy demand will soon overwhelm the minimal cost of single prompts.

Our commitment is to normalize efficient habits now. Just as anti-littering campaigns in the 1960s and 70s transformed littering from a nuisance into a social taboo, small, collective behavioral shifts lead to massive systemic change. By adopting the New Yorker Way—being clear, fast, and direct—all NYU Steinhardt participants become thoughtful contributors to organizational efficiency and global sustainability.

Stop the Small Talk and “Cut the Fluff”

• Avoid greetings & closings. Skip ”Hello,” “Thanks,” “Could you please”. The model doesn't need these pleasantries, and they only accumulate tokens which are wasted energy.
  • Wasteful: “Hi there! Could you kindly summarize this document for me? Thanks so much!”
  • The Efficient New Yorker: “Summarize this document. 5 bullet points.”
• Do not confirm understanding unless required. Avoid responses like “Yep,” “Got it,” “That’s correct,” “I understand,” “Oh, I see.” Let the model assume you are ready for the next instruction.

Talk Fast, Talk Clear: You’ve Got Places To Be!

• Combine instructions. If you have multiple steps, send them in a single, numbered prompt rather than multiple back-and-forth messages. Edit your prompt before sending. This prevents the model from processing multiple, incomplete thought streams.
  • Wasteful: "Write a short email. I want it to be very professional. Oh, and make sure it’s under 100 words." (Three separate prompts/thoughts)
  • The Efficient New Yorker: "Draft a professional email, 75 words max, explaining the Q3 results to the team." (One clear prompt)

Don’t Block the Door! Move!

• Move to a new chat for new topics. Every conversation has a “context window” (the model’s memory of the chat). The longer the chat, the more tokens are consumed with every new prompt just to recall the history.
• Prune the topic. If a chat is getting long and you’re moving into a sub-topic, summarize the necessary historical details and move to a new, fresh chat window.

Fix the Pothole and Move On!

• If it doesn’t need to be done by AI, then don’t use it! Similarly, if you can walk to your destination, then consider walking over driving. Only use AI if it’s something that will benefit your learning or your work that you cannot facilitate through other simpler and more environmentally friendly means.
• Use the “Regenerate” function sparingly. If the first response is 75% good, edit it yourself. Asking the model to “try again” forces it to re-process the entire prompt and previous output, consuming DOUBLE the resources.
• Refine, don’t restart. Instead of asking for a full redo, use specific refinement commands.
  • Wasteful: "I didn't really like that one. Can you please write a new version of the summary? And make sure it's much shorter this time."
  • The Efficient New Yorker: "Make that 100 words shorter and change the tone to professional."

Don’t Waste the Dime! Pick Up the Phone!

• Engage with a human. If you need to brainstorm or process an emotional or sensitive topic, turn off the chat and connect with a friend, a peer, a colleague, a therapist, or a professor. Engage with AI when the task is clear, specific, and transactional. The most sustainable AI interaction is the one that doesn’t happen.

Leadership in the Next Chapter of Innovation

As leaders in education, the arts, and human development, we must view the rise of generative AI not merely as a tool for efficiency, but as a critical subject demanding ethical stewardship. The position that we should simply "not use" this technology is incongruous with the changing global and economic landscape within which our students must graduate and work. Instead, we must be cognizant stewards of efficient and ethical AI use. 

We urge all NYU Steinhardt staff, faculty, researchers, and students to consider the profound impact of AI on their specific fields, their work, and the environment as they embark on this next chapter of technological innovation. Proceeding mindfully and with care, which includes adopting resource-conscious habits, ensures we embody the social responsibility inherent to our disciplines. 

Lasting change in how this technology is shaped requires collective engagement from all of us, spanning regulation, critical research, targeted activism (like Keshaun Pearson's work with the Memphis Community Against Pollution), and local civic involvement (Isaac, 2024). Furthermore, we urge faculty and staff to initiate and foster dialogues about AI's ethical and environmental impact across departments and disciplines.

Take a look at these discussion resources to support your conversations