Let us know about free updates
Can artificial intelligence stop the burning of planets?
There is both good and very good news about the answer to this question.
But as this column is written at the end of yet another swelling UK heatwave, let’s start with a more fun idea that will help AI quickly reduce the trivial chunks of global transport, electricity and food emissions each year.
The new paper from a research team led by British economist Nicholas Stern says it’s worth reading in several counts.
At first, everything that reduces emissions is important, as the transportation, electricity and food sectors account for about half of the world’s emissions.
Researchers believe that AI systems can invade some of these sectors by increasing the use of renewable energy in the power grid. Identifies proteins that make lab-grown meat even tasty. Make electric cars more affordable (using cheaper batteries) and desirable (by predicting the best charging sites).
This could amount to an annual emission reduction of 3.2-5.4 billion tonnes by 2035. This means that AI could reduce emissions in these regions by up to 25%. Even the 3.2 billion tonnes cut out out estimates of electricity-hungry AI and data center emissions, researchers say.
All of these findings are not theoretical. Google’s Deepmind Artificial Intelligence team says the technology can already increase the energy value of wind farms by around 20% and reduce the energy of cooling Google’s data centers by up to 40%.
Deepmind’s Nobel Award-Recognition Alphafold model helped researchers predict the structure of millions of proteins in breakthroughs that could accelerate the use of meat options.
Also, unlike previous research that sought to quantify prospects saving AI planets, Stern was peer-reviewed and not done by Microsoft, Google, or other companies that produce AI products.
Stern led the Stern review commissioned by the UK government in 2006. This indicates that by demonstrating the benefits of early robust climate behavior, it far outweighs the economic costs of not acting. He remains the voice of influential climate policy, even if the world did not act on his discoveries at something like the pace and scale it was necessary.
It’s also very good news about AI and climate change.
Point 1: AI climate technology that looks good in the lab can be a pain in real life. Scientists argued this week that the meta-research project raised false hopes about sucking carbon out of the air, claiming that the approach lacked scientific rigor.
Point 2: AI may be suitable for the other side. The energy company says the technology, in the words of Saudi Arabia Mako, the world’s largest oil company, “makes a real difference” to its operations.
This is unlikely to stop due to a larger obstacle to using AI for the climate: money.
Designing AI systems for companies like Saudi Aramco could have a lot of commercial implications for today’s high-tech companies. As AI researchers like Jack Kelly know, doing it for grid operators in emerging markets may not be the case even if it does more for the planet.
Kelly, in his words, is “fear” about climate change, and was an early discussion with the UK’s nationwide grid, an engineer at Google Deepmind in 2017, when the group was using AI to maximize renewable energy use.
“The first few meetings were really exciting,” Kelly told me. “I felt like I could do something really interesting.”
Alas, this effort was abandoned for reasons that remained unknown. Deepmind declined to comment on reports that there may have been disagreements regarding intellectual property ownership. UK energy operators did not respond to deadlines.
Kelly has left Open Climate Fix, co-founded with Deepmind, a nonprofit that develops AI systems to reduce energy emissions.
Its solar forecasting technology is being used by electricity system operators in Rajasthan, India and the UK. Kelly says it helps reduce emissions by allowing operators to schedule less gas production. His group has other jobs that promise to support the planet, but as he says, “a lot of things have to be done right for that scenario to unfold.”
pilita.clark@ft.com
