Climate change is one of the most pressing challenges of our time, and combating disinformation is crucial for effective action. In a world where large language models (LLMs) like ChatGPT and Google’s Bard have faced concerns over the accuracy of the information they provide, the Endowment for Climate Intelligence (ECI) has taken an innovative approach. They have unveiled ClimateGPT, the first-ever open-source ensemble of AI models dedicated to addressing the impacts of climate change1.
The Need for Trust and Transparency
ClimateGPT aims to provide decision-makers with climate data they can trust. Unlike other LLMs, ClimateGPT draws on scientific data using a robust model that authenticates, secures, and governs the information provided. The model is even integrated with Hedera, a public blockchain, ensuring the highest standards of data integrity1.
ECI, formed by three startups, is on a mission to make reliable climate science information quickly and freely available to researchers, policymakers, and business leaders1. According to Jonathan Dotan, the founder of EQTY Lab and a founding member of ECI, ClimateGPT is part of a broader industry shift toward open-source AI, which emphasizes openness, security, and accountability1.
The Power of Ensemble Models
ClimateGPT is not just one chatbot; it is an ensemble of three task-specific LLMs, each fine-tuned on Meta’s Llama-2 large language model. These models vary in terms of the number of parameters they were trained on, which affects their accuracy. The 70B version, trained on a massive amount of data, outperforms the less accurate 7B model1.
The training data for ClimateGPT comes from various sources, including the Dutch startup Erasmus.AI’s planetary scale corpora. This corpus is derived from over 10 billion web pages and millions of open-access academic articles from across the natural, social, and economic sciences. ClimateGPT’s skills were honed further using data from the Club of Rome’s Earth4All climate science initiative and the UN Sustainable Development Goals (SDGs)1.
Countering Misinformation with Holistic Understanding
One of the key strengths of ClimateGPT is its ability to synthesize interdisciplinary research and break silos to provide a holistic understanding of the impacts of climate change1. Thanks to software from AppTek, ClimateGPT is available in 20 languages, making it accessible to a wide range of users1.
ClimateGPT has also undergone rigorous testing and controls to ensure that it does not spread climate misinformation1. The creators of ClimateGPT have made it available on the Hugging Face community platform, where users can download the model, access its research paper, and explore its training lifecycle using the AI lineage explorer1.
Unleashing the Potential for Positive Change
As ClimateGPT scales up and gets fine-tuned, it has the potential to help researchers, policymakers, and business leaders make informed decisions in a time where climate action is needed more than ever before1. This open-source AI platform is a game-changer, enabling access to reliable climate science information and empowering individuals and organizations to take meaningful action against climate change.