The Carbon Footprint of AI is Already Upon Us: Competitiveness vs. Sustainability?
İstanbul Robot Müzesi Electronic Waste (Photo credit: Katalin Feher)
In contrast, the CO2 emissions by digital or AI services are more invisible or less measurable. According to Melissa Heikkilä in MIT Technology Review, “There is no standardized way to measure carbon dioxide emissions” even if there are available external estimates, such as “OpenAI’s GPT-3 and Meta’s OPT were estimated to emit more than 500 and 75 metric tons of carbon dioxide, respectively, during training”.
How many people or organization have ever checked their carbon footprint when using AI? We have got to be heavy users of several social media, MarTech or FinTech applications driven by AI, and the issues in sustainability are rarely in focus in this process. Not even if power-hungry algorithms already control the ecosystem of online services in a dynamically growing interconnectedness and massive data supply. Services become even more dependent on AI technology, thus, CO2 emissions increase. This is not sustainable at the speed of current technological development.
Yet, what is it that has already begun? There are motivations to measure the related high carbon footprint and expand green AI using renewable energy. The key recommendation is to move the training sessions of machines to a location supplied mainly by renewable sources. However, this engineering approach may not be sufficient for all sustainability issues. Additional efforts are needed from the practices of different industries.
Accordingly, the carbon footprint is calculated chiefly by those who 1) see the limits of technology in the context of competitiveness and 2) think responsibly about the future, in which the carbon footprint will have an exposure role. As a professor at Stanford, Dan Jurafsky formulated “We have to be mindful of whether the benefits of these heavy-compute models are worth the cost of the impact on the environment.”
Marketing, advertising, communication, entertainment or media are responsible for this process as they drive consumption intensively. If there is no other purpose than consumption, the consumer does not become responsible in this process. We are just amazed at possibilities like ChatGPT, Dall-E, MetaHuman or other tools of Generative AI services. However, these also send a wake-up call. According to the article written by Josh Saul and Dina Bass on the Bloomberg platform, “Every new chatbot and image generator requires a lot of electricity, which means the technology may be responsible for a massive and growing amount of planet-warming carbon emissions.”
What conscious decisions are necessary for the creative and cultural industry? The first step is a review what is happening now to face staggering numbers. A great example is “Green Data Dash” to monitor the changing numbers in real-time. The second step is to present the best practices of sustainable info-communication and AI-driven content industry for the future. Carbon.Crane with its products has already made excellent progress in this field. Although AI may work for sustainability on the fly, if not, we will already pay a higher price for the correction during the flight.
Source: Carbon.Crane (https://carboncrane.io/blog?theme=light&lang=en)
#sustainability #carbonfootprint #carbon.crane
