<\/p>\n
Artificial intelligence is consuming enormous amounts of electricity in the US. In accordance with the International Energy Agency, AI systems and data centers used about 415 terawatt hours of power in 2024. That accounts for greater than 10% of the country’s total electricity production, and demand is projected to double by 2030.<\/p>\n
This rapid growth has raised concerns about sustainability. In response, researchers at a School of Engineering have created a proof-of-concept AI system designed to be much more efficient. Their approach could reduce energy use by as much as 100 times while also improving performance on tasks.<\/p>\n
A Hybrid Approach Called Neuro-Symbolic AI<\/strong><\/p>\n The research comes from the laboratory of Matthias Scheutz, Karol Family Applied Technology Professor. His team is developing neuro-symbolic AI, which mixes traditional neural networks with symbolic reasoning. This method mirrors how people approach problems by breaking them into steps and categories.<\/p>\n The work will likely be presented on the International Conference of Robotics and Automation in Vienna in May and can appear within the conference proceedings.<\/p>\n Teaching Robots to See, Understand, and Act<\/strong><\/p>\n Unlike familiar large language models (LLMs) similar to ChatGPT and Gemini, the team focuses on AI systems utilized in robotics. These systems are often called visual-language-action (VLA) models. They extend LLM capabilities by incorporating vision and physical movement.<\/p>\n VLA models absorb visual data from cameras and directions from language, then translate that information into real-world actions. For instance, they’ll control a robot’s wheels, arms, or fingers to finish a task.<\/p>\n Why Traditional AI Struggles With Easy Tasks<\/strong><\/p>\n Conventional VLA systems rely heavily on data and trial-and-error learning. If a robot is asked to stack blocks right into a tower, it must first analyze the scene, discover each block, and determine how you can place them appropriately.<\/p>\n This process often results in mistakes. Shadows may confuse the system a couple of block’s shape, or the robot may place pieces incorrectly, causing the structure to collapse.<\/p>\n These errors are much like the issues seen in LLMs. Just as robots can misplace blocks, chatbots can generate false or misleading outputs. Examples include fabricating legal cases or producing images with unrealistic details similar to extra fingers.<\/p>\n How Symbolic Reasoning Improves Accuracy and Efficiency<\/strong><\/p>\n Symbolic reasoning offers a distinct strategy. As a substitute of relying only on patterns from data, it uses rules and abstract concepts similar to shape and balance. This permits the system to plan more effectively and avoid unnecessary trial and error.<\/p>\n “Like an LLM, VLA models act on statistical results from large training sets of comparable scenarios, but that may result in errors,” said Scheutz. “A neuro-symbolic VLA can apply rules that limit the quantity of trial and error during learning and get to an answer much faster. Not only does it complete the duty much faster, however the time spent on training the system is significantly reduced.”<\/p>\n Strong Leads to Puzzle Tests<\/strong><\/p>\n The researchers tested their system using the Tower of Hanoi puzzle, a classic problem that requires careful planning.<\/p>\n The neuro-symbolic VLA achieved a 95% success rate, compared with just 34% for normal systems. When given a more complex version of the puzzle that it had not encountered before, the hybrid system still succeeded 78% of the time. Traditional models failed every attempt.<\/p>\n Training time also dropped sharply. The brand new system learned the duty in just 34 minutes, while conventional models required greater than a day and a half.<\/p>\n Massive Energy Savings in Training and Use<\/strong><\/p>\n Energy consumption was reduced dramatically as well. Training the neuro-symbolic model required only one% of the energy utilized by a normal VLA system. During operation, it used just 5% of the energy needed by conventional approaches.<\/p>\n Scheutz compared this inefficiency to on a regular basis AI tools. “These systems are only attempting to predict the subsequent word or motion in a sequence, but that may be imperfect, and so they can provide you with inaccurate results or hallucinations. Their energy expense is usually disproportionate to the duty. For instance, while you search on Google, the AI summary at the highest of the page consumes as much as 100 times more energy than the generation of the web site listings.”<\/p>\n The Growing Strain of AI on Power Infrastructure<\/strong><\/p>\n As AI adoption accelerates across industries, demand for computing power continues to climb. Firms are constructing increasingly large data centers, a few of which require tons of of megawatts of electricity. That level of consumption can exceed the needs of entire small cities.<\/p>\n This trend has sparked a race to expand infrastructure, raising concerns about long-term energy limits.<\/p>\n A More Sustainable Path for AI<\/strong><\/p>\n The researchers suggest that current approaches based on LLMs and VLAs might not be sustainable in the long term. While these systems are powerful, they eat large amounts of energy and might still produce unreliable results.<\/p>\n In contrast, neuro-symbolic AI offers a distinct direction. By combining learning with structured reasoning, it could provide a more efficient and dependable foundation for future AI systems.<\/p>\n<\/div>\n <\/p>\n","protected":false},"excerpt":{"rendered":" Artificial intelligence is consuming enormous amounts of electricity in the US. In accordance with the International Energy Agency, AI systems and data centers used about 415 terawatt hours of power in 2024. That accounts for greater than 10% of the country’s total electricity production, and demand is projected to double by 2030. This rapid growth […]<\/p>\n","protected":false},"author":1,"featured_media":314249,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[10],"tags":[46064,397,10554,2943,4262,939],"class_list":["post-314248","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-technology","tag-100x","tag-accuracy","tag-boosting","tag-breakthrough","tag-cuts","tag-energy"],"aioseo_notices":[],"_links":{"self":[{"href":"https:\/\/ebiztoday.news\/index.php\/wp-json\/wp\/v2\/posts\/314248","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/ebiztoday.news\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/ebiztoday.news\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/ebiztoday.news\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/ebiztoday.news\/index.php\/wp-json\/wp\/v2\/comments?post=314248"}],"version-history":[{"count":2,"href":"https:\/\/ebiztoday.news\/index.php\/wp-json\/wp\/v2\/posts\/314248\/revisions"}],"predecessor-version":[{"id":314251,"href":"https:\/\/ebiztoday.news\/index.php\/wp-json\/wp\/v2\/posts\/314248\/revisions\/314251"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/ebiztoday.news\/index.php\/wp-json\/wp\/v2\/media\/314249"}],"wp:attachment":[{"href":"https:\/\/ebiztoday.news\/index.php\/wp-json\/wp\/v2\/media?parent=314248"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/ebiztoday.news\/index.php\/wp-json\/wp\/v2\/categories?post=314248"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/ebiztoday.news\/index.php\/wp-json\/wp\/v2\/tags?post=314248"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}