background image

AI has a large and growing carbon footprint, but there are potential solutions on the horizon

author image

By illuminem briefings

· 2 min read


illuminem summarizes for you the essential news of the day. Read the full piece on The Conversation or enjoy below:

🗞️ Driving the news: AI's potential to tackle the climate crisis is significant, yet its large energy requirements, notably during the training and inference phases of large language models like GPT-3, contribute to carbon emissions
• Innovative approaches such as spiking neural networks (SNNs) and lifelong learning (L2) technologies are being developed to reduce AI's carbon footprint

🔭 The context: The energy demands of AI models, particularly those involving artificial neural networks (ANNs), have skyrocketed, making AI a notable contributor to CO₂ emissions
• ANNs, which are fundamental to most AI systems, require extensive computing power and memory due to their complex calculations
• In contrast, SNNs offer an energy-efficient alternative by mimicking the brain's energy-efficient processing through the use of spikes for information transmission, leading to potentially lower energy requirements and emissions

🌍 Why it matters for the planet: The advancement of SNNs and L2 technologies represents a promising direction in making AI more sustainable
• By significantly reducing the energy consumption of AI systems, these technologies can help mitigate the environmental impact of AI's expanding carbon footprint, contributing to efforts against the climate crisis

⏭️ What's next: Ongoing research and development in SNNs and L2 algorithms aim to bring these technologies closer to practical application, offering a pathway to more energy-efficient AI systems
• This progress, alongside other advancements like quantum computing, holds the potential to reduce the environmental impact of AI significantly

💬 One quote: "SNNs can be up to 280 times more energy efficient than ANNs," highlights the significant potential for energy savings in AI operations.

📈 One stat: Training GPT-3 generated 502 metric tonnes of carbon, equivalent to the annual emissions from 112 petrol-powered cars.

Click for more news covering the latest on climate change

Did you enjoy this illuminem voice? Support us by sharing this article!
author photo

About the author

illuminem's editorial team - providing you with concise summaries of the most important sustainability news of the day.

Follow us on Linkedin, Twitter​ & Instagram

Other illuminem Voices


Related Posts


You cannot miss it!

Weekly. Free. Your Top 10 Sustainability & Energy Posts.

You can unsubscribe at any time (read our privacy policy)