Experts in Germany are changing the way computers think, with the goal of speeding up results and limiting power consumption.
The work was detailed in a news release from the Technical University of Munich and in a scientific journal. It could impact computing for everyday experiences, including self-driving cars and smartphones.
At issue is artificial intelligence and the vast amount of energy that's needed to run the speedy processors. Data centers could use up to 12% of American electricity by 2028, according to the U.S. Energy Department. Meanwhile, Goldman Sachs forecast that AI will drive a 165% increase in computing power demand by the end of the decade.
In Munich, the team thinks that switching computer logic from iterative to probability-based is a solution. Iterative thinking uses algorithms and step-by-step instructions to meticulously work through problems. Munich's approach is "based on the targeted use of values at critical locations" and probability, according to the release.
The inner workings are based on the human brain, using artificial neural networks to pass information to interconnected nodes, or human-made neurons. Importantly, the Munich probability model showed comparable accuracy to iterative processes. But the new approach operated 100 times faster, according to the release.
"Our method makes it possible to determine the required parameters with minimal computing power," physics and enhanced machine learning professor Felix Dietrich said in the statement. "This can make the training of neural networks much faster and, as a result, more energy efficient."
The latter point is crucial. That's because Reuters reported that about 70% of global internet traffic runs through an area near Washington, D.C., called Data Center Alley. It's home to processor sites for Google, Amazon, and other companies.
Reuters detailed how it's difficult for grid operators to manage the massive demand, especially if the centers unexpectedly shift to on-site generators temporarily, leaving a glut of electricity on the grid that unbalances the system.
A lower and consistent power demand can help. This would also mean that less juice is needed from strained power plants, limiting production of heat-trapping air pollution. These gases are linked by NASA to increased severe storm risks, while utility line-dropping extreme weather hurts our electricity system's reliability even more.
Meta and other tech companies are investing in cleaner energy like solar and even nuclear, though the latter form isn't without critics. Meanwhile, big tech companies continue to rely on dirty-fuel-based power sources to run their operations, making consumer advocacy for cleaner policy follow-through important.
TCD Picks » Upway Spotlight
💡Upway makes it easy to find discounts of up to 60% on premium e-bike brands
|
Do you think the federal government should give us tax breaks to improve our homes?
Click your choice to see results and speak your mind. |
Definitely 💯 Speaking up can help prevent companies from touting planet-friendly goals while providing minimal results — otherwise known as greenwashing.
For the Munich team's part, the experts think they are on the cusp of better computing that can save tons of energy.
"We have seen that the accuracy of the new method is comparable to that of iteratively trained networks," Dietrich said.
Using less power to accomplish the same tasks is a goal that can be applied at home, as well.
Updating your property with modern, smart appliances can save you cash and energy. Tax breaks remain for power-saving heat pumps and other tech, but even switching out your old light bulbs for new LED versions can bank hundreds of dollars a year. LEDs also produce five times less pollution compared to traditional illuminators.
Join our free newsletter for weekly updates on the latest innovations improving our lives and shaping our future, and don't miss this cool list of easy ways to help yourself while helping the planet.
