5 ways you can reduce AI’s carbon footprint – don’t be polite!

The rise of artificial intelligence is undeniable. From chatbots assisting with everyday tasks to complex AI systems automating business processes, AI’s influence is rapidly expanding.

However, this growth comes with an environmental cost. Recent reports highlight how even seemingly innocuous habits, like using polite language with AI, can significantly increase energy consumption.

As millions embrace AI, the cumulative energy demand and associated carbon emissions are becoming a serious concern. Technology expert Jan Čurn, founder of Apify, believes that addressing AI’s carbon footprint requires a multi-faceted approach, involving changes in user behaviour, platform development and organizational practices.

Here are five key ways, according to Čurn, to reduce AI’s environmental impact:

1. Optimize Prompt Efficiency:

Čurn points out that seemingly harmless phrases like “please,” “thank you,” and “I hope you’re well” do not contribute to the AI model’s output. However, they increase the “token count” – the way AI processes language – by as much as 30%. While individually small, these extra words, when multiplied across billions of daily prompts, lead to a substantial increase in energy consumption and millions of dollars in unnecessary costs. The solution? Users should strive for concise, direct prompts, focusing on the essential information needed for the AI to perform the desired task.

2. Consider the Impact of Complex Interactions:

Beyond simple queries, users are increasingly engaging in complex, multi-layered conversations with AI. These include scenario-based prompts, simulated interviews and coaching sessions, all of which demand more intricate and lengthy responses from the AI. Compared to straightforward questions, these advanced interactions require significantly more computational resources, further exacerbating the energy consumption problem.

3. Minimize Non-Essential Language:

It’s not just about politeness. Čurn argues that other non-essential elements in prompts, such as filler words, excessive punctuation, and nonverbal symbols like emojis and ellipses, also contribute to a higher token count. He suggests that future AI platforms could incorporate automated prompt optimization tools to streamline user input before processing. This, he believes, could reduce the environmental impact of everyday AI interactions without sacrificing the quality of the model’s output.

4. Develop More Efficient AI Platforms:

The responsibility for reducing AI’s carbon footprint doesn’t solely lie with the end-user. Čurn emphasizes the role of AI platform providers in building sustainability into their systems. He proposes features like prompt efficiency indicators, low-energy or “eco” modes, and automated alerts that warn users when their input exceeds recommended length or complexity thresholds.

5. Promote Organizational Awareness and AI Sustainability Practices:

While AI is rapidly being integrated into business operations, many organizations haven’t yet considered the environmental consequences of large-scale AI usage. Čurn advocates for companies to adopt internal AI sustainability guidelines, promoting efficient prompt design and responsible usage practices.

This is particularly crucial for teams that heavily rely on large language models in high-volume environments like workflow automation and data analysis, where seemingly minor inefficiencies can accumulate into significant energy waste.

Apify 

Google unveils new AI plans and Gemini features at I/O conference