This AI tool for net-zero research discloses its energy use
ChatNetZero 3.0 draws on research from the Net Zero Tracker, the most comprehensive database of such commitments. Read More
- The tool was developed by the group that publishes Net Zero Tracker, along with climate researchers from the University of North Carolina and tech company Arboretica.
- ChatNetZero is updated frequently with new data from the core tracker.
- The resource uses less energy than ChatGPT to answer the same question.
ChatNetZero, an AI chatbot trained to handle questions about greenhouse gas emissions targets, can now tell you how much energy was used to generate the answer.
The tool, first introduced in October 2023, was developed by the research groups that publish the widely used Net Zero Tracker, along with the Data-Driven Envirolab at the University of North Carolina at Chapel Hill and tech company Arboretica.
The group created this tool after noticing that general AI chatbots aren’t a reliable source of answers to questions about greenhouse gas emissions reduction targets.
“It fills a niche that doesn’t exist in the world of generic AI,” said Angel Hsu, director of DataDriven Envirolab. “More people are starting to ask these questions. How do we ensure that the answers can be trusted?”
ChatNetZero directly links to the Net Zero Tracker, which keeps tabs on pledges by countries, more than 2,000 of the world’s largest companies and cities with populations of more than 500,000 people.
Corporate sustainability professionals can use the chatbot to check how their own targets are presented or to gather data about peers and competitors. The update includes new algorithms for analyzing that data, including the ability to parse information from charts that are part of corporate ESG disclosures.
“We trained the model to show facts without asking it to make a judgement about the facts,” said James Zhang, co-developer of ChatNetZero and founder of Arboretica.
Because sustainability teams are increasingly concerned about accounting for the energy used by AI resources, the ChatNetZero developers added a feature that displays the amount of energy the AI uses per query compared with the same question asked of ChatGPT. Because ChatNetZero is trained with a specific data set it requires less power to deliver answers.
For example, if everyone in Ireland asked a net-zero question using ChatNetZero versus ChatGPT, it would save about 1,400 kilowatt-hours of electricity, roughly the amount of power required to run a U.K. household for several months.
Subscribe to Trellis Briefing
