The hidden energy footprint of AI
Interview with Victor Pereboom, co-founder and Chief Technology Officer of UbioOps.
Victor is the co-founder and CTO at UbiOps. Combining an engineering background with a business mindset, he enjoys working with the team at UbiOps to deliver a product to enable data science teams to embed ML & AI into their organisation.
From observing AI’s operational process, Victor has observed the high energy consumption of machine learning models. Many of the strategies and techniques are very compute intensive and, therefore, will consume even more energy than before. Furthermore, as AI develops, it will most likely need even more complicated procedures — this will require more power, leading to more impact on the environment.
Victor argues that the environmental impact of digitalisation is very hidden. Many companies are deploying AI models at scale, but more awareness is needed to make it more environmentally friendly. Furthermore, users don’t realise how our gadgets are impacting the environment. We are all aware of how we can be environmentally friendly in recycling plastic, but what about technology? It is estimated that a single Google search could power a low-energy light bulb for an hour — what needs to change?
In our interview with Victor, we’ll be discussing how we can measure and rate the environmental impacts of AI models.
Stay tuned!