Digital product studio ustwo and the University of Bristol have launched Prism, an open-source plug-in that estimates the carbon emissions generated by developers’ use of AI tools. The tool is being promoted by the Green Software Foundation community.
The plug-in sits inside a developer’s integrated development environment (IDE) and its makers claim it can provide real-time, directional estimates of the energy and carbon associated with AI interactions. The aim is to let engineering teams consider emissions alongside cost, speed and performance while building software.
Prism converts the number of AI tokens used into estimated energy consumption and carbon emissions. Its methodology draws on guidance from the Green Software Foundation and recent academic research.
Prism team, ustwo
“AI has become an everyday part of software development, but its environmental cost remains largely hidden,” said Paolo Rizzi, sustainability principal at ustwo, in a statement.
The tool is not intended to provide precise carbon accounting. Instead, it gives developers an indication of the relative environmental impact of their AI usage at the point where technical decisions are being made.
Most existing sustainability tools focus on infrastructure or company-wide emissions. Prism instead targets individual development workflows, where choices about models, prompts and AI usage take place.
“Even directional estimates can be valuable,” said Nick Hegarty, ustwo’s executive director of technology.
Prism was developed with students from the University of Bristol and is being released free and open source. ustwo said the project’s methodology and code will remain open to scrutiny and contributions from developers and sustainability researchers.
The launch comes as companies increasingly embed generative AI into software development, while the energy demands associated with training and running AI models receive greater scrutiny.
(Image: ChatGPT)
