SogetiLabs gathers distinguished technology leaders from around the Sogeti world. It is an initiative explaining not how IT works, but what IT means for business.
4 thoughts on “Green IT – Adding Sustainability to Your Operational Excellence”
Great article!
First comment: Machine Learning (especially training neural networks) are one of the fastest growing sinners leading to very large CO2 emissions. So choose carefully where machine learning makes sense and where an easily calculated heuristic would do.
Second comment: Look here for an approach to reduce your organisation’s IT related carbon footprint – https://www.capgemini.com/2020/07/how-to-create-a-green-it-virtuous-circle/
@Erik that is a great remark. I believe neural nets are an amazing invention, but like blockchain their potential and power consumption tightly follow the exponential curve of GPU advancements. It’s undeniable that even bad ML beats the best heuristics more and more often, so the effort cannot be solely put on carefully studying all possible options. Do you have any proposals to make machine learning greener ?
At the moment I don’t have any ideas to make ML greener.
To clarify – heuristics is not about studying all possible solutions – it’s about a human guessing the algorithm based on a small data sample rather than having ML guessing the algorithm based on a huge data sample. Sometimes we human can do a much better job 🙂 And, we have the ability to argue why we created the algorithm the way we did and hence how the results were derived – that’s still only an ambition to create explainable AI.
Many tools and applications could be conceived on very low tech & low resolution screen, maybe even persistent e-ink. But no major companies seem to go in this direction for now.
Great article!
First comment: Machine Learning (especially training neural networks) are one of the fastest growing sinners leading to very large CO2 emissions. So choose carefully where machine learning makes sense and where an easily calculated heuristic would do.
Second comment: Look here for an approach to reduce your organisation’s IT related carbon footprint – https://www.capgemini.com/2020/07/how-to-create-a-green-it-virtuous-circle/
@Erik that is a great remark. I believe neural nets are an amazing invention, but like blockchain their potential and power consumption tightly follow the exponential curve of GPU advancements. It’s undeniable that even bad ML beats the best heuristics more and more often, so the effort cannot be solely put on carefully studying all possible options. Do you have any proposals to make machine learning greener ?
At the moment I don’t have any ideas to make ML greener.
To clarify – heuristics is not about studying all possible solutions – it’s about a human guessing the algorithm based on a small data sample rather than having ML guessing the algorithm based on a huge data sample. Sometimes we human can do a much better job 🙂 And, we have the ability to argue why we created the algorithm the way we did and hence how the results were derived – that’s still only an ambition to create explainable AI.
Many tools and applications could be conceived on very low tech & low resolution screen, maybe even persistent e-ink. But no major companies seem to go in this direction for now.