Skip to main content

Integrating Sustainability into Responsible AI Frameworks: The Time is Now

Addressing the sustainability challenges of Gen AI requires a multifaceted approach.

Amir Sayegh

By Amir Sayegh

August 21, 2024

•

2 minute read

Data servers with blue lights

This is part of a Geotab series on Responsible AI, where we unpack the implications of Generative AI (Gen AI). It leverages recent work on generative AI, spanning the launch of Geotab Ace and the accompanying Responsible AI Whitepaper as well as our Gen AI Maturity Index.

 

In 2022, an obscure San Francisco startup quietly unveiled ChatGPT-3. This unassuming release was the beginning of a transformative new era in AI, set to reshape the world. In a short time Gen AI has touched every industry offering unprecedented capabilities in content creation, language processing, and problem-solving. However, it comes at a significant environmental cost as demand for computational resources skyrockets raising serious sustainability concerns.  

The Resource Demands of Gen AI

The surge in Gen AI applications has the potential to transform industries, enhance productivity, and create new avenues for innovation. However, training and operating Gen AI models requires vast amounts of data and computational power to train and operate (inference), often involving large-scale data centers with high-performance hardware. As models grow more complex, their resource demands increase, contributing to a larger carbon footprint. Ignoring its impact is no longer an option. Let’s explore: 

  • CO2 Emissions: Training large AI models can produce around 626,000 pounds of CO2, equivalent to approximately 300 round-trip flights between New York and San Francisco​
  • Daily Emissions: BLOOM, a large language model, emits greenhouse gasses comparable to driving 49 miles in an average gas-powered car (19 kilograms of CO2 per day of use).
  • Water Consumption: Each session with GPT-3 (10 to 50 responses) can consume the equivalent of a 500ml bottle of water​. Gen AI's energy consumption is expected to be 10 times higher in 2026 than in 2023. 
  • Water Withdrawal: Global AI demand may account for 4.2 - 6.6 billion cubic meters of water withdrawal in 2027, more than half of the UK’s total annual water withdrawal.

At Geotab, our priority is incorporating Sustainable AI, which is essential for responsible AI deployment. Sustainable AI means using AI systems in ways that align with sustainable business practices. 

Innovative Solutions for Sustainable AI

Addressing the sustainability challenges of Gen AI requires a multifaceted approach. Several areas to explore include: 

  • Efficient Chip Architectures: Develop/Adopt new chip designs to enhance processing efficiency and use less energy.
  • Efficient, Task-Specific Models: Create AI models that prioritize efficiency without sacrificing performance and are tailored for specific tasks by optimizing their algorithms and structures. This approach reduces the need for heavy computing and saves energy without sacrificing performance. 
  • Smaller Models: Design purpose-built, smaller AI models for specific tasks to cut down on computational requirements and energy usage.
  • Mixing Models: Use a mixture of models for targeted processing, reducing computational load.
  • Edge Computing: Deploy AI models closer to data sources to minimize data transfer and reduce energy use.
  • Sustainable Data Centers: Invest in renewable energy and energy-efficient technologies for data centers.
  • Green Tech Partners: Choose technology partners committed to sustainability to promote a greener tech ecosystem.
  • Advanced Cooling Solutions: Use modern cooling technologies to save water and reduce energy consumption.
  • Smart Resource Allocation: Use dynamic resource allocation to optimize computational resources and reduce waste.
  • Optimized Data Management: Use efficient data storage solutions and representative subsets of data to minimize the volume processed.
  • Advocacy and Offsetting: Raise awareness about the environmental impact of AI and advocate for sustainable practices within the AI community. Invest in carbon offset projects and seek green certifications to support sustainable initiatives.

Embracing a Sustainable Future

Organizations using AI must think sustainably, and include sustainability practices in their responsible AI frameworks, like those mentioned above, that progress on ESG and sustainability targets. AI itself can also be used to help reduce its footprint and aid in fighting the effects of climate change. At Geotab, Geotab Ace is being leveraged by fleet managers to make better decisions and enhance fleet sustainability.

 

As Gen AI continues to shape our future, it is imperative that we work together to balance innovation with sustainability, and ensure a greener, more responsible future for all. 

Subscribe to get industry tips and insights


Amir Sayegh
Amir Sayegh

Amir Sayegh is the Associate Vice President of Data Product Discovery at Geotab

Subscribe to get industry tips and insights

View last rendered: 12/14/2024 02:24:29