TL;DR:
- ClimateGPT, an open-source AI platform, is launched to address the challenges of climate change.
- The model offers researchers, policymakers, and business leaders valuable insights and decision-making support.
- Users can access the model, research paper, and AI lineage explorer on Hugging Face.
- Benchmark scores demonstrate exceptional efficiency in climate-specific tasks.
- Erasmus.AI developed ClimateGPT using a vast corpus of web and academic content.
- AppTek’s AI specialists contributed to climate-specific instruction fine-tuning.
- ClimateGPT leverages an advanced cryptographic framework for security and governance.
- The platform was introduced during COP28 and is now accessible globally through green energy data centers.
- Enterprises can use the Retrieval-Augmented Generation (RAG) AI platform for real-time applications.
- Several corporate entities, including TAQA, Etihad Rail, ADNEC Group, and Masdar, have adopted ClimateGPT.
- Transparent and responsible AI governance is a priority, with blockchain and IPFS-based solutions.
- The model incorporates marginalized groups’ perspectives to address climate change.
Main AI News:
The Endowment for Climate Intelligence (ECI) is thrilled to announce the launch of ClimateGPT, a pioneering open source ensemble of AI models dedicated to addressing the rapidly evolving challenges posed by climate change. ClimateGPT is poised to empower researchers, policymakers, and business leaders with the tools they need to make well-informed decisions in an era of climate uncertainty.
Now available on Hugging Face, ClimateGPT offers users the opportunity to access the model, its comprehensive research paper, and utilize a groundbreaking AI lineage explorer, providing unprecedented visibility into the training lifecycle of ClimateGPT. This release underscores ECI’s unwavering commitment to the open science and open source AI communities, marking a significant stride towards responsible AI development. Notably, the model’s benchmark scores demonstrate a remarkable 10x efficiency improvement on climate-specific tasks and an innovative cascading machine translation capability, reclaiming nearly 94% of fluency performance compared to native multilingual language models. This achievement signifies a profound milestone in creating an equitable and rigorously audited AI model, meticulously fine-tuned by experts representing a diverse range of perspectives and local stakeholders.
A Closer Look at the Model
Following over four years of rigorous research, testing, and development of more than 100 Large Language Models, Erasmus.AI has meticulously crafted the ClimateGPT corpus from its expansive planetary scale dataset—one of the world’s most extensive collections of web and academic content, featuring insights on climate, extreme weather, the Club of Rome’s Earth4All, and UN Sustainable Development Goals (SDGs). The Erasmus corpus draws from an extensive pool of over 10 billion web pages and millions of open-access academic articles, enabling ClimateGPT to synthesize interdisciplinary research and bridge silos, forming a holistic understanding of climate change’s impacts across natural, social, and economic sciences.
In collaboration with AppTek’s AI and language research scientists, the model underwent specialized climate-specific instruction fine-tuning (IFT), resulting in a dataset and benchmark that facilitates access to knowledge spanning scientific disciplines across more than 20 languages. EQTY Lab played a pivotal role in partnership with Further Ventures to design the ClimateGPT platform, which leverages an advanced cryptographic framework for authentication, security, and governance, ensuring the responsible deployment of AI models.
AI Powered by Renewable Energy
The launch of ClimateGPT’s initial node took place during COP28 at Abu Dhabi’s Al Dhafra Solar PV, a 2-gigawatt facility and the world’s largest single-site solar plant. Access to the model has now expanded to Microsoft’s green energy data centers worldwide. Notably, ClimateGPT was trained using 256 Nvidia H100s, renowned for their energy efficiency and powered by hydroelectric energy in Puyallup, Washington.
Designed for Enterprise
ClimateGPT offers groundbreaking capabilities through the Retrieval-Augmented Generation (RAG) AI platform, enabling enterprises to harness the open source model for real-time applications with proprietary datasets. Integrations with Salesforce and Databricks facilitate rapid deployment and fine-tuning of the model. At the UN COP28 Conference, several corporate entities, including TAQA, Etihad Rail, ADNEC Group, and Masdar, a renewable energy company, proudly announced their adoption of the platform.
Ensuring Responsible AI
Transparency and governance are at the core of ClimateGPT’s principles. To achieve this, the platform relies on an innovative, trusted AI solution from EQTY Lab that records the entire AI lifecycle on the Hedera enterprise-grade blockchain and stores model data on Protocol Lab’s IPFS and the Filecoin protocols. Responsible AI initiatives have been initiated in collaboration with experts from the open source, trust and safety, and climate misinformation communities to establish robust guardrails for the model’s deployment. Additionally, the ECI has partnered with Khalifa University and inclusion consultants RSSC to develop an instruction fine-tuning (IFT) pilot, incorporating the perspectives of marginalized groups most affected by climate change.
In the words of Daniel Erasmus, CEO of Erasmus.AI, “This breakthrough goes beyond technological advancement. It’s about accelerating our collective journey towards a sustainable future. Policymakers, business leaders, and researchers now have a powerful decision support platform to bring us closer to that future.”
Christian Dugast, PhD, Chief Scientist for LLMs at AppTek, emphasizes the model’s grounding in objective facts and comprehensive training: “We ensure that our answers are firmly rooted in domain-specific documents, supporting both completion and citation mechanisms to provide well-summarized responses and retrieve relevant documents.“
Ariana Fowler, Head of Research at EQTY Lab, highlights the urgency for responsible AI in addressing global challenges: “ClimateGPT serves as a beacon of innovation, contributing to open source research and offering enterprise-ready solutions. This launch demonstrates the potential for diverse stakeholders with a shared mission to introduce new public AI utilities, emphasizing the collaborative nature of our collective climate intelligence.”
ClimateGPT stands as a testament to the power of AI in driving positive change for our planet, offering hope for a more sustainable future through responsible and innovative technology.
Conclusion:
The introduction of ClimateGPT represents a significant milestone in the market for AI-driven solutions to climate change challenges. This open source platform provides invaluable resources for researchers, policymakers, and businesses, fostering informed decision-making and innovative climate solutions. With its impressive efficiency and comprehensive training, ClimateGPT has the potential to expedite the transition to a more sustainable future, making it a crucial asset for enterprises seeking to address climate-related issues. Furthermore, its commitment to transparency and responsible AI sets a standard for ethical AI development in the industry, reinforcing the importance of accountable technology in addressing global challenges.
