Artificial Intelligence (AI) has become an integral part of modern technology, yet its future is contingent on resolving a critical energy crisis. The continued advancement of AI is not a guarantee unless an energy breakthrough occurs that would allow us to meet its growing demand for electricity.
The carbon footprint of AI is almost on par with Bitcoin, consuming energy equivalent to entire developed nations. Science Alert reported last year that the Information Technology industry's contribution to global CO2 emissions is around 2%, a figure expected to grow rapidly. Consulting firm Gartner predicts the AI sector alone will account for 3.5% of global electricity by 2030.
The training of GPT-3, a predecessor of ChatGPT, exemplifies this issue. It required approximately 1,287 megawatt hours and 10,000 computer chips, equating to the annual power usage of 121 average American homes. This process also produced around 550 tonnes of carbon dioxide. OpenAI, the developer of ChatGPT, reportedly spends about US$700,000 daily on computing costs to serve its global user base.
Need for carbon-free energy
Sam Altman, CEO of OpenAI, recently addressed this challenge. At a Bloomberg event alongside the World Economic Forum in Davos, he stressed the need for a breakthrough in carbon-free energy production to sustain AI's growth. Altman's investments reflect his commitment, with significant funding in nuclear fusion company Helion Energy and his role as chairman of nuclear microreactor company Oklo.
While AI's energy consumption is a concern, it also plays a crucial role in global decarbonization efforts. AI's capabilities in renewable energy forecasting, smart grid management, and new material development are vital for environmental sustainability. However, as Sasha Luccioni from Hugging Face notes, the environmental footprint of AI must be considered to ensure its responsible use in combating global warming.