{"id":3950,"date":"2025-04-16T16:00:34","date_gmt":"2025-04-16T16:00:34","guid":{"rendered":"https:\/\/nationalgunowner.org\/index.php\/2025\/04\/16\/a-group-of-harvard-researchers-aim-to-demystify-a-i-learning-processs\/"},"modified":"2025-04-16T16:00:37","modified_gmt":"2025-04-16T16:00:37","slug":"a-group-of-harvard-researchers-aim-to-demystify-a-i-learning-processs","status":"publish","type":"post","link":"https:\/\/nationalgunowner.org\/index.php\/2025\/04\/16\/a-group-of-harvard-researchers-aim-to-demystify-a-i-learning-processs\/","title":{"rendered":"A Group of Harvard Researchers Aim to Demystify A.I. Learning Processs"},"content":{"rendered":"
\n
Hidenori Tanaka speaks at NTT Group\u2019s Upgrade 2025 conference in San Francisco on April 9, 2025. Victor Dey for Observer<\/span><\/figcaption><\/figure>\n

For years, A.I. has been considered a \u201cblack box\u201d technology<\/a>, whose inner workings remain largely mysterious even to its creators because the system is supposed to \u201clearn\u201d itself. But a group of Harvard researchers are on a quest to find out exactly how A.I. learns.\u00a0<\/span>At <\/span>a conference in San Francisco last week, Japanese IT services company <\/span>NTT Group<\/span> launched the <\/span>Physics of Artificial Intelligence (PAI) Group<\/span>, based at Harvard University\u2019s Center for Brain Science. Just as physics reveals the laws that govern motion and energy, the group aims to identify the fundamental principles that drive A.I.\u2019s learning and reasoning.<\/span><\/p>\n

\n<\/section>\n

\u201cTo truly evaluate and solve A.I.\u2019s black-box paradox, we need to understand it on a psychological and architectural level\u2014how it perceives, decides and why,\u201d Hidenori Tanaka<\/a>, PAI\u2019s group leader, told Observer at last week\u2019s event. Tanaka is also a lead A.I. researcher at Harvard\u2019s Center for Brain Science.<\/span><\/p>\n

Currently, the capabilities of an A.I. are measured through an approach called benchmarking, which typically involves testing models against a set of standardized tasks or problems, such as answering scientific questions, recognizing images or playing games. Tanaka believes this method is limited. \u201cWe need to go beyond benchmarking. It\u2019s an insult to judge A.I. models based on mere computational power and how well they solve a couple of tough problems. It completely misses the cognitive depth models may have,\u201d he said.<\/p>\n

More importantly, understanding how an A.I. model works better helps humans improve it. \u201cUncovering the root causes of an A.I. model\u2019s initial reasoning behavior can play a key role in minimizing bias and hallucinations in upcoming systems,\u201d Kazu Gomi<\/a>, president and CEO of NTT Research<\/a>, told Observer at last week\u2019s event.<\/p>\n

To achieve this, PAI is building \u201cmodel experimental systems,\u201d or controlled digital experiment environments through which developers can observe how an A.I. model\u2019s learning and reasoning curve evolves over time. By crafting numerous multimodal datasets consisting of images and text across physics, chemistry, biology, math and language topics, the team is probing how A.I. interprets previously unseen concepts and retains information.\u00a0PAI plans to partner with A.I. developers worldwide to improve these datasets through insights gleaned from real-world experiments.\u00a0<\/span><\/p>\n

\u201cThe goal is to give A.I. systems a structured playground. Unlike internet-scraped data, these are intentionally crafted datasets, with each data point entry serving a distinct, predefined function,\u201d explained Tanaka. \u201cJust like medications act on specific neurons to treat a physical condition in humans, we\u2019re looking at how information triggers responses within A.I. models at the neural or node level.\u201d<\/span><\/p>\n

Tanaka has a Ph.D. in theoretical physics from Harvard. His research focuses on studying<\/span> the brain\u2019s computational principles and aligning them with physical laws. <\/span><\/p>\n

PAI is a spinoff from NTT Research\u2019s Physics and Informatics Lab, established in 2019, where Tanaka previously led a group researching machine learning and sustainable data center solutions. The group has delivered over 150 papers in A.I. research. PAI\u2019s previous research on <\/span>neural network pruning algorithms<\/span><\/a> has been cited over 750 times.<\/span><\/p>\n

PAI\u2019s core team includes multiple Harvard researchers. It also collaborates with Harvard neuroscientist Venkatesh Murthy, Princeton professor and former NTT Research scientist Gautam Reddy, and Stanford\u2019s Surya Ganguli. Ganguli is a senior researcher at Stanford\u2019s Institute for Human-Centered Artificial Intelligence (HAI), led by the A.I. pioneer Fei-Fei Li<\/a>, and has co-authored several papers with Tanaka.\u00a0<\/span><\/p>\n

\t\t\t\t\"A<\/p><\/div>\n