An academic has calculated the numbers to show how the energy demand of artificial intelligence (AI) could be equivalent to the energy demand of an entire country.
Thank you for reading this post, don't forget to subscribe!Advertisement
Recent growth in the capabilities and applications of artificial intelligence (AI) has opened up many debates about the potential shortcomings of the technology; From students cheating in exams to machines replacing humans in the workplace, and even the threat of complete destruction of the human race.
Now another potential pitfall of AI may be on the horizon, with experts warning that the computing power required to run the necessary algorithms and machine learning processes could contribute to climate change due to the amount of energy it uses. Is.
Alex de Vries, an academic who is a PhD candidate at the Vrije Universiteit Amsterdam, has suggested that if AI were used in every Google search for a year, the equivalent of electricity used to power a small country like Ireland. Quantity will be required.
I am writing a comment magazine joulea companion publication room“Given the increasing demand for AI services, it is very likely that AI-related energy consumption will increase significantly in the coming years,” he says.
De Vries is the founder of DigiconomistA website highlighting the unintended consequences of digital trends.
AI has emerged as a growing and important digital trend in 2023, and they warned that when it becomes more widely adopted, its energy demands could exceed what some countries can handle.
Generative AI is being used by the general public every day, with chatbots like OpenAI’s ChatGPT and MidJourney’s image creation tool being the most popular. To generate their outputs, the models on which they are built require the process of machine learning, where they are fed large amounts of data.
In his commentary, de Vries explains how Hugging Face, a US-based AI company, said its multilingual text-generation AI used about 433 megawatt-hours (MWh) during its training, which is about the same as 40 average homes. Enough to provide electricity. US for one year
It’s not just training that uses a lot of energy.
Efficient AI means more demand
De Vries’ analysis shows that when a tool like ChatGPT outputs text based on signals, it uses “a significant amount of computing power and thus energy”. He says that running ChatGPT can consume 564 MWh of electricity per day.
While developers are working to make their AI tools more efficient, De Vries says this could result in a phenomenon called the Jevons Paradox.
“Making these devices more efficient and accessible could result in us having more applications and allowing more people to use them,” he said.
De Vries estimates that, based on available data regarding power consumption and AI, if Google were to use AI for its approximately 9 billion daily searches, it would require 29.2 terawatt hours (TWh) of electricity each year. Will be – equal to the annual electricity consumption. Of Ireland.
“The potential increase highlights that we need to be very careful about what we use AI for,” De Vries said. “It’s energy intensive, so we don’t want to put it into all kinds of things where we don’t really need it,” he said.
While the above scenario is unlikely to occur in the short term, AI’s ability to process demand is bound to increase. De Vries estimates that worldwide AI-related electricity consumption could increase by 85 to 134 TWh annually by 2027, depending on AI server production projections.
This would be equivalent to the annual electricity needs of countries like the Netherlands, Argentina and Sweden.
De Vries has previously covered the energy costs of another high-profile technology, which requires using huge amounts of electricity for cryptocurrency transactions.
Some estimates suggest that Bitcoin, the most popular cryptocurrency, could emit the same amount of carbon dioxide each year as the entire New Zealand.