story: Machine Learning reduced cooling energy consumption in Google's data centres by 40%, in a way humans had not imagined.
A data centre is a facility housing computer servers that are used for all kinds of purposes. With large numbers of high-performance servers, temperatures in such spaces can rise significantly, so efficient cooling is essential, and A.I. discovered that the secret was in the air!
By analysing historical data from thousands of sensors inside data centres, Google DeepMind trained and optimised neural networks, resulting in a significant reduction in energy consumption. It did this by making intelligent adjustments to airflow inside the data centres. It directed cooled air to specific hotspots where it was needed most and optimised airflow in a way people had not thought of. In addition, A.I. also adjusted fan speeds so that every corner of the data centre received the right amount of cooling without wasting energy.
Even more remarkable is that Google DeepMind originated in learning to play old games such as Breakout, Pong and Space Invaders. It made headlines when DeepMind defeated world champion Lee Sedol in the complex board game Go.
This highlights the particular strength of A.I. in pattern recognition, where human operators fall short or are unable to implement solutions because of the complexity and scale of the available data. Where the human eye and brain have limits in processing large amounts of complex information, A.I. excels at discovering hidden patterns and suggesting new solutions. This ability to derive refined insights from enormous datasets also offers countless possibilities in many other sectors.
✉ Feel free to connect with me or send me a message if you are interested in exploring the possibilities of AI!
🖼️ Image generated by ChatGPT 4 / Dall-E 3, based on the text above.
