AI workloads: Driving sustainable IT

Summary

John Frey, HPE's Chief Technologist, discusses the importance of sustainable AI solutions. He emphasizes HPE's commitment to net-zero carbon emissions by 2040 and highlights five ethical AI principles: privacy, human focus, inclusivity, robustness, and responsibility. Frey introduces five levers of IT sustainability—data, software, equipment, energy, and resource efficiency—that are crucial for AI lifecycle management. He stresses the need for efficiency in AI, from data collection to inference, and invites audience engagement for continuous improvement.

Takeaways

• 🌿 **Sustainability Commitment**: HPE has a science-based net-zero target to reduce carbon emissions to zero by 2040, emphasizing the importance of sustainability in AI solutions.
• 🤖 **AI Ethical Principles**: HPE adheres to five ethical principles for AI: Privacy, Human Focus, Inclusivity, Robustness, and Responsibility.
• 🔥 **Increased Power Density**: AI workloads can significantly increase power density, with some racks requiring over 50 kilowatts, necessitating more efficient cooling solutions.
• 🌡️ **Heat Management**: As power densities rise, heat loading increases, and traditional cooling methods may no longer suffice, prompting a shift towards liquid cooling technologies.
• 💡 **Efficiency Focus**: AI solutions should be designed with efficiency in mind, considering the entire lifecycle from hardware to application to data.
• 📊 **Data Efficiency**: Optimizing data sets for AI is crucial to avoid unnecessary work and to ensure that AI models are trained on clean and relevant data.
• 💻 **Software Efficiency**: Applications and algorithms should be optimized for efficiency to reduce the computational resources required for AI workloads.
• 🔌 **Equipment Efficiency**: Utilizing hardware infrastructure to its full potential can significantly improve the efficiency of AI workloads.
• ⚡ **Energy Efficiency**: Maximizing performance per watt is essential for both environmental and operational reasons, as power constraints and carbon footprint considerations become more critical.
• 🔄 **Resource Efficiency**: Beyond hardware, considering the efficiency of cooling, power conversion, and other resources is vital for sustainable AI operations.
• 🔄 **Lifecycle Considerations**: AI solutions should be evaluated throughout their lifecycle, from development to deployment to retirement, to ensure sustainability.

Q & A

What is John Frey's role at HPE?

John Frey is HPE's Chief Technologist and Director of Sustainable Transformation, leading their IT Sustainability practice.

What is HPE's commitment regarding carbon emissions?

HPE has a science-based net zero target, committing to reduce their carbon emissions across their value chain to zero by 2040.

What percentage of HPE's carbon footprint is attributed to customer use of their technology?

Two-thirds of HPE's carbon footprint is attributed to the use of their technology by their customers.

What are the five ethical principles for AI according to HPE?

HPE's five ethical principles for AI are: Privacy enabled and secure, Human focused, Inclusive, Robust, and Responsible.

How does HPE ensure that their AI solutions adhere to ethical principles?

HPE has an advisory group that analyzes their AI solutions for the five ethical elements to ensure they are in place before bringing solutions to market.

What is the significance of increased power density in AI workloads?

Increased power density in AI workloads, which can reach up to 50 plus kilowatts per rack, requires more power in a small space and leads to higher heat loading.

Why is liquid cooling technology becoming more important for AI sustainability?

Liquid cooling technology is becoming more important due to the increased power densities and heat loading in AI workloads, which necessitate more efficient cooling solutions.

What are the five levers of IT sustainability that HPE has developed?

HPE's five levers of IT sustainability are: Data efficiency, Software efficiency, Equipment efficiency, Energy efficiency, and Resource efficiency.

How does HPE's approach to AI sustainability apply across different infrastructures?

HPE's approach to AI sustainability applies across all infrastructures, whether in public cloud, data centers, co-lo, or at the edge, by focusing on the five levers of IT sustainability.

What is the importance of considering the full lifecycle impacts in AI sustainability?

Considering the full lifecycle impacts in AI sustainability is important for understanding the long-term efficiency, carbon footprint, and overall sustainability of AI solutions from design to disposal.

How does HPE encourage engagement and feedback from customers on their sustainability resources?

HPE encourages engagement and feedback by making their white papers and resources freely available online and inviting customers to provide input and feedback to improve and refresh the materials every two years.