Artificial intelligence needs a lot of power, which will only grow as companies rush to make it bigger, smarter, and more complicated. A big carbon footprint is a tough problem for a business that prides itself on being a powerful way to save the planet.
But Sam Altman, the head of OpenAI, the company that made ChatGPT, says there is a clear way out of this tough situation: nuclear Fusion.
Altman has put hundreds of millions of dollars into Fusion. In recent comments, he has said that he thinks this futuristic technology, seen by many as the holy grail of clean energy, will one day provide the huge power that next-generation AI will need.
Altman said in a January interview, “We need Fusion and other renewable energy sources to be scaled up. Without a solution, we can’t get there. When computer scientist and podcaster Lex Fridman asked Altman in March how to solve AI’s “energy puzzle,” Altman said the same thing.
The sun and other stars get their power from nuclear Fusion, which is still likely decades away from being fully understood and used in everyday life on Earth. Some experts think that Altman’s focus on a future energy breakthrough shows that the AI industry hasn’t devised a plan for how to meet AI’s rising energy needs soon.
They say AI can help them solve a major issue in their quest for almost endless clean energy.
Alex de Vries, a data scientist and researcher at Vrije Universiteit Amsterdam, believes it matches the notion that people are typically “wishful thinking” about what they can do to improve the environment. In an interview with CNN, he remarked, “It would be much smarter if we concentrated on what we have now and whatever we can do now than to keep hoping for something that might happen.”
A spokesperson for OpenAI only talked about Altman’s words in January and on Fridman’s podcast when CNN referred them to questions.
The AI business is interested in nuclear Fusion. Fusion is the process of smashing two or more atoms together to make one heavier. This gives off a lot of energy.
It doesn’t release carbon dioxide into the air or leave behind long-lasting nuclear waste. It’s a tempting idea for a clean, safe, and plentiful energy source.
“Recreating the conditions in the centre of the sun on Earth is a huge challenge,” according to Aneeqa Khan and her team of researchers at the University of Manchester in the United Kingdom, and they predict that technology won’t be ready until the second half of the century.
Fusion cannot address the climate catastrophe at this time, Khan told CNN. We must employ existing low-carbon technologies, such as fission and renewable energy, soon.”
Fission is the main way that nuclear energy is made
The problem is getting enough clean energy to meet AI’s growing needs right now instead of using fossil fuels, which warm the Earth. It’s especially hard because the push to electric everything from cars to heating systems is making more clean energy needed.
The International Energy Agency recently did a study that said the energy used by data centres, cryptocurrencies, and AI could double in the next two years. The IEA said the sector was responsible for about 2% of the world’s energy needs in 2022.
The study said that between 2023 and 2026, demand for AI will grow at least 10 times faster than supply.
Along with the energy needed to make chips and other hardware, AI needs a lot of computing power to “train” models by giving them huge amounts of data and then using that training to answer a user’s question.
Companies are trying to add the technology to apps and online searches as it improves, increasing the need for computing power. In a recent study on AI’s energy footprint, de Vries found that an AI-powered online search could use at least 10 times more power than a normal search.
“Bigger is better when it comes to AI,” de Vries said, which has companies building big models that use a lot of energy. He said this is the main issue with AI because the idea that bigger is better is at odds with sustainability.
Mike Khoo, head of the climate disinformation program at Friends of the Earth and co-author of a report on AI and climate, said that things are especially bad in the US, where energy demand is going up for the first time in about 15 years. He spoke with CNN about the country’s diminishing electricity supplies.
The increase in data centres adds to the rising energy demand. According to Boston Consulting Group research, data centres would require three times as much energy as they do by 2030, enough to power about forty million homes in the United States.
Khoo emphasised the impending difficult decisions surrounding electricity distribution, whether it is allocated to data centres powering powerful AI or to numerous households. He emphasised that energy distribution should not only benefit the wealthy.
AI companies frequently overlook two crucial factors when concerned about their energy consumption: First of all, AI can assist in resolving the climate issue.
AI is a powerful tool for sustainability solutions
People are already using the tech to track pollution, map forests, predict the weather, and monitor ice melting. A recent study by Boston Consulting Group, whGooglehiredogle, said AI could help cut down on up to 10% of the pollution that heats the planet.
AI might also be able to help move nuclear Fusion forward. In February, scientists at Princeton said they had found a way to use the technology to predict when nuclear fusion processes might become unstable. This was a big step toward making the technology available to the public.
AI companies also say they are working hard to make things run more smoothly. Google says its data centres work 1.5 times better than the average business data centre.
On Saturday, a partial curfew was in place so evacuees could move around and rescue equipment could be moved. According to the authorities, this was because of a series of fires that had never happened before.
Scientists have warned that the world has exceeded a dangerous temperature limit for a year.
A Microsoft spokeswoman stated that the company is allocating resources to examine AI technology’s energy consumption and carbon footprint. Simultaneously, the organisation is investigating ways to improve large-scale systems’ efficiency, including training and practical applications.
A “tremendous” development in AI’s practicality, says de Vries
He warned that this doesn’t necessarily mean AI will use less energy.
De Vries also said that the past of technology and automation shows that it might be the reverse. He said something about cryptocurrency. He stated that efficiency advances have never resulted in cryptocurrency mining using less energy. “The demand for some goods and services goes up when we make them more efficient.”
The US has a political push to examine how AI affects the climate more closely. In February, Sen. Ed Markey proposed a bill requiring AI companies to be more open about how they affect the environment, such as by using a lot of electricity in their data centres.
During that time, Markey said, “The creation of the next generation of AI tools cannot come at the cost of the health of our planet.” However, not many people think the bill will get the backing of both parties. It needs to become law.
Khoo said that companies are in an “arms race to produce the next thing,” and the development of AI that is getting more complicated and energy-hungry is seen as inevitable. He also said that this means models that are bigger and bigger and use more and more power.
