{"682905":{"#nid":"682905","#data":{"type":"news","title":"Brain-Inspired AI Breakthrough Spotlighted at Global Conference","body":[{"value":"\u003Cp dir=\u0022ltr\u0022\u003EResearchers at Georgia Tech have taken a critical step forward in creating efficient, useful and brain-like artificial intelligence (AI). The key? A new algorithm that results in neural networks with internal structure more like the human brain.\u003C\/p\u003E\u003Cp dir=\u0022ltr\u0022\u003EThe study, \u201c\u003Ca href=\u0022https:\/\/openreview.net\/forum?id=THqWPzL00e\u0022\u003ETopoNets: High-Performing Vision and Language Models With Brain-Like Topography\u003C\/a\u003E,\u201d was awarded a spotlight at this year\u2019s\u0026nbsp;\u003Ca href=\u0022https:\/\/iclr.cc\/\u0022\u003EInternational Conference on Learning Representations\u003C\/a\u003E (ICLR), a distinction given to only 2 percent of papers. The research was led by graduate student\u0026nbsp;\u003Ca href=\u0022https:\/\/www.murtylab.com\/group\u0022\u003E\u003Cstrong\u003EMayukh Deb\u003C\/strong\u003E\u003C\/a\u003E alongside\u0026nbsp;\u003Ca href=\u0022https:\/\/psychology.gatech.edu\/\u0022\u003ESchool of Psychology\u003C\/a\u003E Assistant Professor\u0026nbsp;\u003Ca href=\u0022https:\/\/www.murtylab.com\/\u0022\u003E\u003Cstrong\u003EApurva Ratan Murty\u003C\/strong\u003E\u003C\/a\u003E.\u003C\/p\u003E\u003Cp dir=\u0022ltr\u0022\u003E\u003Ca href=\u0022https:\/\/sites.gatech.edu\/research\/iclr-2025\/\u0022\u003EThirty-two of Tech\u2019s computing, engineering, and science faculty represented the Institute at ICLR 2025\u003C\/a\u003E, which is globally renowned for sharing cutting-edge research.\u0026nbsp;\u003C\/p\u003E\u003Cp dir=\u0022ltr\u0022\u003E\u201cWe started with this idea because we saw that AI models are unstructured, while brains are exquisitely organized,\u201d says first-author Deb. \u201cOur models with internal structure showed more than a 20 percent boost in efficiency with almost no performance losses. And this is out-of-the-box \u2014 it\u2019s broadly applicable to other models with no extra fine-tuning needed.\u201d\u003C\/p\u003E\u003Cp dir=\u0022ltr\u0022\u003EFor Murty, the research also underscores the importance of a rapidly growing field of research at the intersection of neuroscience and AI. \u201cThere\u0027s a major explosion in understanding intelligence right now,\u201d he says. \u201cThe neuro-AI approach is exciting because it helps emulate human intelligence in machines, making AI more interpretable.\u201d\u003C\/p\u003E\u003Cp dir=\u0022ltr\u0022\u003E\u201cIn addition to advancing AI, this type of research also benefits neuroscience because it informs a fundamental question: Why is our brain organized the way it is?,\u201d Deb adds. \u201cMaking AI more interpretable helps everyone.\u201d\u003C\/p\u003E\u003Ch3\u003E\u003Cstrong\u003EBrain-inspired blueprints\u003C\/strong\u003E\u003C\/h3\u003E\u003Cp dir=\u0022ltr\u0022\u003EIn the brain, neurons form topographic maps: neurons used for comparable tasks are closer together. The researchers applied this concept to AI by organizing how internal components (like artificial neurons) connect and process information.\u0026nbsp;\u003C\/p\u003E\u003Cp dir=\u0022ltr\u0022\u003EThis type of organization has been tried in the past but has been challenging, Murty says. \u201cHistorically, rules constraining how the AI could structure itself often resulted in lower-performing models. We realized that for this type of biophysical constraint, you simply can\u2019t map everything \u2014 you need an algorithmic solution.\u201d\u003C\/p\u003E\u003Cp dir=\u0022ltr\u0022\u003E\u201cOur key insight was an algorithmic trick that gives the same structure as brains without enforcing things that models don\u0027t respond well to,\u201d he adds. \u201cThat breakthrough was what Mayukh (Deb) worked on.\u201d\u0026nbsp;\u003C\/p\u003E\u003Cp dir=\u0022ltr\u0022\u003EThe algorithm, called\u0026nbsp;\u003Ca href=\u0022https:\/\/github.com\/murtylab\/topoloss\u0022\u003ETopoLoss\u003C\/a\u003E, uses a loss function to encourage brain-like organization in artificial neural networks, and it is compatible with many AI systems capable of understanding language and images.\u0026nbsp;\u003C\/p\u003E\u003Cp dir=\u0022ltr\u0022\u003E\u201cThe resulting training method, TopoNets, is very flexible and broadly applicable,\u201d Murty says. \u201cYou can apply it to contemporary models very easily, which is a critical advancement when compared to previous methods.\u201d\u0026nbsp;\u003C\/p\u003E\u003Ch3\u003E\u003Cstrong\u003ENeuro-AI innovations\u003C\/strong\u003E\u003C\/h3\u003E\u003Cp dir=\u0022ltr\u0022\u003EMurty and Deb plan to continue refining and designing brain-inspired AI systems. \u201cAll parts of the brain have some organization \u2014 we want to expand into other domains,\u201d Deb says. \u201cOn the neuroscience side of things, we want to discover new kinds of organization in brains using these topographic systems.\u201d\u003C\/p\u003E\u003Cp dir=\u0022ltr\u0022\u003EDeb also cites possibilities in robotics, especially in situations like space exploration where resources are limited. \u201cImagine running a model inside a robot with limited power,\u201d he says. \u201cStructured models can help us achieve 80 percent of performance with just 20 percent of energy consumption, saving valuable energy and space. This is still experimental, but it\u0027s the direction we are interested in exploring.\u201d\u003C\/p\u003E\u003Cp dir=\u0022ltr\u0022\u003E\u201cThis success highlights the potential of a new approach, designing systems that benefit both neuroscience and AI \u2014 and beyond,\u201d Murty adds. \u201cWe can learn so much from the human brain, and this project shows that brain-inspired systems can help current AI be better. We hope our work stimulates this conversation.\u201d\u003C\/p\u003E","summary":"","format":"limited_html"}],"field_subtitle":"","field_summary":[{"value":"\u003Cp\u003E\u003Cem\u003EResearchers at Georgia Tech have developed an algorithm that helps AI models develop internal organization just like the human brain \u2014 boosting efficiency by 20 percent.\u003C\/em\u003E\u003C\/p\u003E","format":"limited_html"}],"field_summary_sentence":[{"value":"Researchers at Georgia Tech have developed an algorithm that helps AI models develop internal organization just like the human brain \u2014 boosting efficiency by 20 percent."}],"uid":"35599","created_gmt":"2025-06-26 16:33:50","changed_gmt":"2025-06-26 16:46:30","author":"sperrin6","boilerplate_text":"","field_publication":"","field_article_url":"","location":"Atlanta, GA","dateline":{"date":"2025-06-26T00:00:00-04:00","iso_date":"2025-06-26T00:00:00-04:00","tz":"America\/New_York"},"extras":[],"hg_media":{"677290":{"id":"677290","type":"image","title":"Neurons growing in a culture dish (NASA)","body":"\u003Cp\u003ENeurons growing in a culture dish (NASA)\u003C\/p\u003E","created":"1750955780","gmt_created":"2025-06-26 16:36:20","changed":"1750955780","gmt_changed":"2025-06-26 16:36:20","alt":"Neurons growing in a culture dish (NASA)","file":{"fid":"261174","name":"Neurons-in-a-culture-dish.jpg","image_path":"\/sites\/default\/files\/2025\/06\/26\/Neurons-in-a-culture-dish.jpg","image_full_path":"http:\/\/hg.gatech.edu\/\/sites\/default\/files\/2025\/06\/26\/Neurons-in-a-culture-dish.jpg","mime":"image\/jpeg","size":130235,"path_740":"http:\/\/hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/2025\/06\/26\/Neurons-in-a-culture-dish.jpg?itok=LjpPFJsT"}},"677291":{"id":"677291","type":"image","title":"School of Psychology Assistant Professor\u00a0Apurva Ratan Murty","body":"\u003Cp\u003ESchool of Psychology Assistant Professor\u0026nbsp;Apurva Ratan Murty\u003C\/p\u003E","created":"1750955976","gmt_created":"2025-06-26 16:39:36","changed":"1750955976","gmt_changed":"2025-06-26 16:39:36","alt":"School of Psychology Assistant Professor\u00a0Apurva Ratan Murty","file":{"fid":"261175","name":"Ratan.jpg","image_path":"\/sites\/default\/files\/2025\/06\/26\/Ratan.jpg","image_full_path":"http:\/\/hg.gatech.edu\/\/sites\/default\/files\/2025\/06\/26\/Ratan.jpg","mime":"image\/jpeg","size":162869,"path_740":"http:\/\/hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/2025\/06\/26\/Ratan.jpg?itok=1DIDvH7C"}},"677292":{"id":"677292","type":"image","title":"Graduate Student Mayukh Deb","body":"\u003Cp\u003EGraduate Student Mayukh Deb\u003C\/p\u003E","created":"1750956091","gmt_created":"2025-06-26 16:41:31","changed":"1750956091","gmt_changed":"2025-06-26 16:41:31","alt":"Graduate Student Mayukh Deb","file":{"fid":"261176","name":"Deb.jpg","image_path":"\/sites\/default\/files\/2025\/06\/26\/Deb.jpg","image_full_path":"http:\/\/hg.gatech.edu\/\/sites\/default\/files\/2025\/06\/26\/Deb.jpg","mime":"image\/jpeg","size":143409,"path_740":"http:\/\/hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/2025\/06\/26\/Deb.jpg?itok=LxkXF9or"}}},"media_ids":["677290","677291","677292"],"groups":[{"id":"1278","name":"College of Sciences"},{"id":"1188","name":"Research Horizons"},{"id":"443951","name":"School of Psychology"}],"categories":[{"id":"194606","name":"Artificial Intelligence"},{"id":"153","name":"Computer Science\/Information Technology and Security"},{"id":"135","name":"Research"},{"id":"134","name":"Student and Faculty"},{"id":"193158","name":"Student Competition Winners (academic, innovation, and research)"},{"id":"8862","name":"Student Research"}],"keywords":[{"id":"172970","name":"go-neuro"},{"id":"192253","name":"cos-neuro"},{"id":"192863","name":"go-ai"},{"id":"187915","name":"go-researchnews"}],"core_research_areas":[{"id":"193655","name":"Artificial Intelligence at Georgia Tech"},{"id":"193653","name":"Georgia Tech Research Institute"},{"id":"193656","name":"Neuro Next Initiative"}],"news_room_topics":[],"event_categories":[],"invited_audience":[],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[{"value":"\u003Cp\u003EWritten by Selena Langner\u003C\/p\u003E\u003Cp\u003EContact: \u003Ca href=\u0022mailto: jess.hunt@cos.gatech.edu\u0022\u003EJess Hunt-Ralston\u003C\/a\u003E\u003C\/p\u003E","format":"limited_html"}],"email":[],"slides":[],"orientation":[],"userdata":""}}}