Openning Ceremony
The opening ceremony of the summit focused on the key role of artificial intelligence in promoting global sustainable development. A number of key speakers in the field of global sustainable development discussed in depth the opportunities and difficulties of utilizing AI technology to address environmental and social challenges, outlining a blueprint for a green future empowered by technology. This speech not only leads to the theme of the summit, but also provides a new direction of thinking about the potential of AI in environmental protection, social impact and other areas.
Opening Speaker 1: Jerry Huang
At the opening ceremony of this Green AI Summit, Jerry Huang delivered a key speech to introduce the core theme and team mission of the summit.
1. Vision and Mission of Green AI Summit
The main goal of the summit is to promote global sustainable development in a technology-enabled way. He pointed out that AI technology is advancing rapidly, but it is also putting tremendous pressure on energy systems and the environment. Therefore, the team hopes to promote more equitable and sustainable AI applications through research, industry collaboration, academic exchanges and policy development.
2. White paper release and content highlights
He officially released the team's first annual white paper, Global AI Environmental Impact Report. The report analyzes the environmental impacts of data centers and AI technologies in terms of energy consumption, carbon emissions, and water use, covering policy differences and development status in major economies such as the U.S., EU, and China. The report also specifically examines the life cycle assessment of data centers and AI models to reveal their all-encompassing impact on the environment.
3. Establishing a Global AI Environmental Impact Index
The team developed a “Green AI Index” to quantify the environmental impact of AI. The index is based on a multi-layered analysis that provides a comprehensive set of metrics, from the carbon and water footprint of data centers to the lifecycle assessment of AI models, and Jerry emphasized that it is critical to use the data to support regulatory and policy development.
4. Enhancing Industry Collaboration and Knowledge Sharing
Jerry called on academics, policymakers, and industry leaders around the world to join the research and discussion on green AI. He mentioned that the team is creating a journal on AI and sustainability, organizing a sustainability leadership challenge, and engaging the broader public and youth in the field.
5. Driving the future: a green AI community that everyone can join
He emphasized that the Green AI Summit is not just an academic and industry exchange, but also a starting point for building a global green AI community. The team plans to organize regular weekly and annual summits, and will release more white papers in the future to promote the practical application of AI in global sustainable development.
Prof. Junwei Cao from the National Research Center for Information Science and Technology, Tsinghua University, delivered a keynote speech on “Networked Energy Convergence Infrastructure and Artificial Intelligence Applications” at the Green AI Summit. He introduced how the two-way convergence of AI and energy technologies can drive a more sustainable future.
1. Bi-directional contribution of network energy convergence
Artificial Intelligence can be used to develop more advanced energy applications to predict energy consumption, manage distributed energy resources, and improve efficiency through big data and deep learning. Energy technology can likewise support AI development in reverse, providing stable and green energy security for AI computing through network energy convergence.
2. “Energy Internet” and “Energy Router” Architecture
Energy Internet is similar to the Internet architecture, which adopts a bottom-up distributed structure and is connected to the national grid through local energy networks.
The energy router is similar to a network router, which can carry out energy scheduling and exchange between different components, support energy storage (energy cache) and energy management, and construct network energy devices under low voltage.
3. Application Scenarios and Demonstration Projects
Distributed energy: including microgrid, electric vehicle energy storage and demand side management, interconnected through AC/DC networks, and combined with heating, cooling and hydrogen energy systems to form a complete distributed energy ecosystem.
Smart Buildings and Smart Cities: Utilizing AI to enable load forecasting and PV power generation prediction to improve the accuracy of energy management, with demonstration systems in several Chinese cities, including Shenzhen, Chengdu and Shanghai.
New Data Centers: New AI data centers in China will gradually be integrated with new energy infrastructure, enabling the coupling of task scheduling and renewable energy energy management, supporting the optimization of electric vehicle charging strategies, and reducing the burden on the power grid.
4. Future Outlook: National “East Counts, West Counts” Project
With the implementation of China's “East Counts, West Counts” strategy, network-energy convergence infrastructure will play a key role in supporting the construction of AI infrastructure. Resource-rich regions in the west can provide computing power to support data analytics needs in the east, thus realizing the efficient use of energy resources across the country.
Prof. Cao's speech provided a forward-looking perspective on the future development of Green AI, and we look forward to working together to promote low-carbon and environmentally friendly AI with the joint efforts of the global Green AI community.
Opening Speaker 3: Suchi Gopal
In the opening session of the Green AI Summit, Suchi Gopal, Professor of Earth and Environment at Boston University, discussed the strategic siting of green AI, focusing on the balance between renewable energy, water, and sustainability.
1. AI and the carbon footprint of data centers
AI applications such as image generation produce carbon emissions comparable to those produced by driving 4.1 miles. U.S. data centers are projected to consume as much power as the national electricity consumption of Japan or India over the next five years, underscoring the significant environmental impact of the AI revolution.
2. Data Center Location Issues on a Global Scale
The rapid growth of AI has created a demand for data centers, and each country needs its own sovereign AI data center. However, this also poses siting challenges, such as access to resources and the need for electricity and water, especially in regions such as Africa and Asia.
3. Sustainable Development Goals (SDGs) considerations in site selection
She emphasized the UN's Sustainable Development Goals (SDGs), particularly clean water and sanitation, sustainable cities, and affordable energy, which are critical in data center siting. Several countries are facing serious challenges in terms of resources such as water and electricity.
4. Green Energy and Microgrid Needs
Many developing data centers are facing power and water shortages and places like Noida in India have been affected. To cope with these issues, data centers are beginning to explore microgrid solutions such as solar power.
5. Water Competition and Environmental Justice Issues
Data centers have a huge demand for water, with large data centers consuming 1-5 million gallons of water per day, affecting tens of thousands of people in the surrounding area. The fairness and reasonableness of data center locations has also led to environmental justice protests by local residents.
6. Pros and cons of data center development
Data centers can increase tax revenues and improve network connectivity, but they can also create water and land use problems for nearby residents. Therefore, the location of data centers involves “environmental justice” issues.
7. Green Bonds and Financing Support
Her book, “The Green Tech Revolution,” explores the trend of green bonds and bank loans for data center development, and how many countries are beginning to support the construction of green AI data centers through these financial instruments.
Prof. Gopal's presentation prompted participants to think deeply about the environmental impact and resource consumption of the rapid development of AI data centers.
Opening Speaker 4: Zhaohao Ding
Prof. Zhaohao Ding, Professor at the School of Electrical and Electronic Engineering, North China Electric Power University (NCPU), participated in the opening session via online and shared his team's insights on low carbon operation and energy management in data centers.
1. Energy demand challenges in data centers
With the rapid development of AI technologies (e.g., ChatGPT and generative AI), the energy consumption of data centers has increased significantly. How to find a balance between energy consumption and sustainable development has become a key issue in the data center industry.
2. Challenges of utilizing renewable energy
There is a significant mismatch between the energy demand of data centers and the generation of renewable energy in terms of time and geographic location. In order to maximize the use of renewable energy, data centers need to increase the flexibility of their energy usage.
3. Three Data Center Flexibility Strategies
Workload Scheduling: Schedule some insensitive offline tasks (e.g., data backup and machine learning) during the time period when renewable energy supply is sufficient.
Resource Management: Optimize server power consumption by dynamically adjusting server voltage and frequency.
Auxiliary equipment control: Flexibly adjust the temperature of the data center's cooling system to keep it within an acceptable range and reduce unnecessary energy consumption.
4. Industry Practice Examples
Google's Carbon Smart Computing: Google implements workload transfer between data centers to utilize clean energy from different regions.
North China Electric Power University (NCPU) and Alibaba's collaborative project: Cross-provincial power load transfer between data centers in Jiangsu and Hebei provinces to achieve more efficient use of renewable energy.
5. Implementation Challenges
Engineering barriers: Insufficient real-world data and engineering constraints lead to a gap between academic research and real-world applications.
Non-engineering barriers: Different stakeholders have different goals, such as coordination between the IT and energy sectors, and mechanisms are needed to share the resulting synergies to incentivize long-term sustainable collaboration.
Professor Zhaohao Ding pointed out that although industry and academic research on decarbonizing data centers has been going on for more than 10 years, practical applications are still very limited. In the future, overcoming technical and non-technical barriers will be key to achieving the development of low-carbon data centers.
Opening Speaker 5: Jason Mohagheg
Jason Mohaghegh, Associate Professor of Comparative Literature at Babson College and head of the AI Ethics Research Group, brings a unique perspective on the future ethical challenges of superintelligent AI.
1. The Double Imagination of Ancient Civilizations: Paradise and Doomsday
Professor Mohaghegh began with ancient civilizations, noting that ancient civilizations like Babylon, Sumeria, Egypt, and China not only portrayed utopian beginnings of beauty, but were also filled with destructive stories of doom. This imagining of a perfect world alongside a destructive one foreshadows a complex vision of humanity's future.
2. Utopia and Anti-Utopia in AI and Environmental Protection
In the field of AI and environmental protection, there are both rosy visions of technological miracles and concerns about potential disasters. Superintelligent AI may bring about ecological and technological breakthroughs, but it is also accompanied by the risk of extinction and environmental crises, forming a picture of a future where “hope and fear” coexist.
3. Experiments in envisioning the city of the future: the fusion of utopia and anti-utopia
In his lab at Babson College, Prof. Mohaghegh leads students in designing hypothetical cities of the future, encouraging them to explore the dual possibilities of utopia and anti-utopia through bold experimental interpretations of cutting-edge innovations and trends. This experiment is not only a technological innovation, but also a cultural and ethical experiment.
4. AI's “cognitive gap”: the barrier between human and superintelligent understanding
The development of superintelligent AIs has brought about an intellectual gap - that is, humans may not be able to understand the behaviors and ways of thinking of superintelligence. For example, some scholars have suggested that the “fear of death” should be programmed into AI systems in order to equip them with human survival anxiety.
5. AI metaphors: elves, oracles, and magical objects
Different cultures and civilizations understand AI in terms of their own metaphors:
Genies: have the ability to fulfill human wishes, but often end badly.
Oracles: can reveal the truth, but are always presented in riddles and difficult language.
Magical objects: have miraculous powers, but can be misused or fall into the wrong hands. Each of these metaphors demonstrates a potential double-sidedness of AI technology.
6. “Going Too Far”: Ancient Warnings and Modern Ethical Considerations
Every culture has the expression “going too far”, and this ancient warning is particularly important in the development of modern technology. How do we define “going too far”, and is the development of AI beyond the boundaries of our understanding? Prof. Mohaghegh emphasized that these are questions that deserve deeper reflection, and that the real way out lies in facing, exploring, and accepting the unknown future.
Associate Professor Mohaghegh's talk revealed the deep challenges of AI ethics, encouraging us to understand the future of technology from historical, cultural and psychological perspectives, while warning us of the great risks that may be hidden behind technological advances.
Opening Speaker 6: Amal Aldaej
Ms. Amal Aldaej, Senior Advisor for Global Engagement at the National Center for Vegetation Cover and Combating Desertification in Saudi Arabia, delivered a fascinating online presentation. She shared innovative ways on how AI can drive global environmental resilience, improve land management and foster international cooperation.
1. The important role of AI in combating desertification and land restoration
AI has great potential to improve environmental research and help global communities develop sustainably. Particularly in combating desertification and improving land, generative AI can process massive amounts of data, predict environmental changes and design adaptive solutions.
At the National Center for Vegetation Cover and Combating Desertification (NCVC) in Saudi Arabia, AI is used to build models that identify early signs of land degradation and drive more effective land management strategies.
2. AI-driven ecosystem monitoring and restoration
AI monitors vegetation cover in real time, responds to desertification-threatened areas in a timely manner, and selects suitable tree species by analyzing local climatic conditions to ensure the success of new afforestation projects and improve carbon sink capacity.
3. Global cooperation on AI environmental applications
Different countries and organizations are already enhancing environmental resilience through AI. For example, Africa uses AI to analyze drone imagery data to protect endangered species; Australia uses AI to optimize forest fire management; and the United Nations Environment Programme (UNEP) develops AI tools to monitor deforestation.
International cooperation and data sharing have become key factors for AI to drive transnational environmental projects, such as the G20's Global Tree Planting Initiative, which uses deep learning techniques with multi-temporal satellite imagery to analyze vegetation dynamics.
4. Strategies and future visions to promote AI for environmental protection
In order to better promote the application of AI in environmental protection, she emphasized the development of a clear strategic roadmap and operational model, and the construction of collaborative mechanisms at the national and global levels.
She also invited participation in the forthcoming sixteenth Conference of the Parties to the United Nations Convention to Combat Desertification (COP16) in Riyadh, through which it was hoped that innovation-driven discussions on land restoration and desertification control would be fostered, in particular by encouraging the active participation of the private sector.
Opening Speaker 7: Paulo Carvao
Paulo Carvao, Senior Fellow at the Mossawa Rahmani Center for Business and Government at Harvard Kennedy School, shared the energy consumption pressures that current large-scale AI models put on the environment and proposed three innovative directions to achieve green AI. His presentation triggered the participants to think about the future development of AI, especially on the balance between energy consumption and sustainability.
1. High energy consumption and environmental burden of large AI models
Training large AI models such as GPT-3 requires huge energy consumption. For example, the training process of GPT-3 consumed 1,287 MWh of electricity and emitted 502 tons of carbon dioxide, which is equivalent to the carbon emissions of 112 gasoline cars in a year or the annual electricity consumption of 120 U.S. households. This demonstrates the huge environmental impact of AI model training and the lack of transparency in the field.
In addition, AI consumes significant amounts of energy during the reasoning process, for example a single query consumes ten times more energy than a Google search. Data centers require large amounts of water in addition to electricity, consuming thousands of gallons per day, further burdening the environment.
2. The rise of nuclear energy: a new strategy for enterprises to cope with AI energy demand
As AI's energy needs increase, many large corporations are turning to nuclear energy to ensure the sustainability of their energy supply. Microsoft is exploring Small Modular Reactors (SMRs) and working with energy companies to revitalize the Three Mile Island nuclear facility in Pennsylvania, U.S. AWS has also signed a nuclear energy agreement to meet its goal of achieving net-zero carbon emissions by 2040.
By introducing nuclear energy into their data center energy systems, these companies hope to achieve a green transformation in AI energy consumption and move toward a more sustainable future.
3. Three Innovative Directions for Achieving Green AI
AI model innovation: Optimizing AI model architectures, algorithms, and training methods to reduce energy consumption. For example, Liquid Neural Networks (LNN) can continuously adapt to streaming data, consume less computing resources, and improve transparency.
Data center innovation: Reduce data center energy consumption by optimizing cooling systems, adopting efficient hardware, and utilizing waste heat and renewable energy.
Data center location innovation: Reduce carbon footprint by locating data centers in areas with abundant hydropower, solar or wind energy. For example, Phoenix, Arizona, USA is becoming a new data center hub due to abundant land and clean energy.
4. Government and business working together to promote green AI
Governments and businesses need to work together to achieve green AI. governments should set targets for efficient energy use and renewable energy to drive the green transformation of data centers. In addition, governments should support green technology R&D and incentivize private sector innovation.
Enterprises should practice green practices throughout the life cycle of AI technologies, establish transparent environmental impact reporting mechanisms, and responsibly manage the lifelong environmental impacts of their technologies.