Last week, AIWS Weekly wrote about the potential of Grover, an AI program to detect fake news. This is possible by recognizing the trace or signature that is left as artifacts of the AI, such as OpenAI’s GPT-2 text generator, that produces the fake text. These programs are examples of the AI machines that require computationally intensive training of very large neural networks. These “energy-guzzling” machines may fuel global warming.
It is suggested that the carbon footprint of training a single AI could be five times the lifetime emissions of an average car; New Scientist reported. “From an energy perspective, and from a carbon reduction perspective, we should be thinking about designing the services and making sure the algorithms are efficient as possible,” says Chris Priest at the University of Bristol. Big firms, in the like of Amazon and Google, are already taking major steps for more uses of renewable energy in training their AIs.
Not long ago, we heard similar criticisms, albeit not the same target. “Bitcoin will burn the planet down”. Déjà vu. As we are heading toward a green future, living in a green economy, no matter what advances a technology could offer, it seems to always be “it is energy efficiency, stupid.”
