Researchers call for transparency on AI tools' carbon footprint; training GPT-3 took an estimated 1.287 GWh of electricity, or what ~120 US homes use in a year
Bloomberg : Tweets: @zachweinberg , @dinabass , @izzyz , @sjmmcd , @brianjjackson , @doctorow , @mattmday , @environment , @katekayereports , @nic__carter , @tante , @bigblackjacobin , and @bloombergasia Tweets: @zachweinberg : No. No we don't. We do not hold back technical progress that will benefit literally the entire world because of wanting to understand its carbon footprint. https://twitter.com/... Dina Bass / @dinabass : We need a lot more transparency on the carbon emissions of large AI products and models like ChatGPT, researchers told @joshfromalaska and me. https://www.bloomberg.com/... Israel Gonzalez-Brooks / @izzyz : We need to weigh deceleration efforts against the opportunity cost in terms of lost productivity, health and resources for the world's impoverished. Artificial intelligence is the only technology single-handedly reducing the intractability of UN sustainable development goals. https://twitter.com/... Samuel Miller McDonald / @sjmmcd : 1) AI is not “progress” & certainly does not/will not “benefit literally the entire world” 2) Superfluous, harmful tech should be forcibly stopped from being developed 3) Preventing CO2 emissions & ecological harm should be prerequisites for any new/existing tech https://twitter.com/... Brian Jackson / @brianjjackson : Focusing on the way power is consumed rather than how it's produced seems like an unproductive way to go about decarbonization to me. https://twitter.com/... Cory Doctorow / @doctorow : Now is not the time to ask questions! The time to ask questions is after I get rich and it's too late to do anything! https://twitter.com/... Matt Day / @mattmday : “Governments and companies may decide that using GPT-3 or other large models for researching cancer cures or preserving indigenous languages is worth the electricity and emissions, but writing rejected Seinfeld scripts or finding Waldo is not.” https://www.bloomberg.com/... @environment : The creation of every new chatbot and image generator requires a lot of electricity, which means artificial intelligence may be responsible for a massive and growing amount of planet-warming carbon emissions. https://benvgo.com/R7Hh226 Kate Kaye / @katekayereports : Researchers like @SashaMTL want transparency on AI power use & emissions. Great to see this coverage from @joshfromalaska @dinabass. AI's carbon problem is not just an LLM thing. It ties directly to data use issues we're analyzing at @privacyforum. https://www.bloomberg.com/... Nic Carter / @nic__carter : can't say you weren't warned https://www.bloomberg.com/... @tante : “AI” as machine learning is a basically a fossil fuel technology. It eats a lot of energy in a time where we don't have enough renewable sources of it leading to more fossil fuel burning. Sure, it can be worth it but choosing to be blind to that fact is infantile. https://twitter.com/... Edward Ongweso Jr / @bigblackjacobin : The idea that every single technological development is a step forward which will benefit humanity is as stupid and childish as the caricature conjured up that insists we should freeze all tech for carbon emissions. We should not mindlessly develop tech based on empty promises. https://twitter.com/... @bloombergasia : Every new chatbot and image generator requires a lot of electricity, but no one knows exactly how much. https://www.bloomberg.com/...