{"id":90892,"date":"2025-11-11T18:40:51","date_gmt":"2025-11-11T15:40:51","guid":{"rendered":"https:\/\/forklog.com\/en\/?p=90892"},"modified":"2025-11-11T18:45:55","modified_gmt":"2025-11-11T15:45:55","slug":"stanford-scientist-identifies-physical-thinking-as-ais-main-obstacle","status":"publish","type":"post","link":"https:\/\/forklog.com\/en\/stanford-scientist-identifies-physical-thinking-as-ais-main-obstacle\/","title":{"rendered":"Stanford Scientist Identifies Physical Thinking as AI&#8217;s Main Obstacle"},"content":{"rendered":"<p>Artificial intelligence is not yet fully capable of understanding the physical world. This remains the primary challenge for the technology, <a href=\"https:\/\/drfeifei.substack.com\/p\/from-words-to-worlds-spatial-intelligence\">stated<\/a> Stanford University computer science professor Fei-Fei Li.<\/p>\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>\u201cLeading AI technologies like large language models (LLM) have changed how we access and work with abstract knowledge. However, they remain masters only in words: eloquent but inexperienced, knowledgeable yet unsubstantiated,\u201d he believes.\u00a0<\/p>\n<\/blockquote>\n<p>According to the scientist, the emergence of \u201cspatial intelligence\u201d will transform how people \u201ccreate and interact with real and virtual worlds, revolutionising literature, art, robotics, science, and more.\u201d<\/p>\n<p>Developing such technology requires training models not only on \u201clanguage\u201d but also on the physical properties of the world.\u00a0<\/p>\n<p>Li asserts that artificial intelligence is rapidly approaching the limits of text-based learning, and ultimately its progress will depend on \u201cworld models\u201d\u2014a new type of generative AI that must tackle a fundamentally different set of tasks than LLMs.<\/p>\n<blockquote class=\"twitter-tweet\">\n<p lang=\"en\" dir=\"ltr\">AI\u2019s next frontier is Spatial Intelligence, a technology that will turn seeing into reasoning, perception into action, and imagination into creation. But what is it? Why does it matter? How do we build it? And how can we use it?<\/p>\n<p>Today, I want to share with you my thoughts on\u2026 <a href=\"https:\/\/t.co\/L0bnJcCUqc\">pic.twitter.com\/L0bnJcCUqc<\/a><\/p>\n<p>\u2014 Fei-Fei Li (@drfeifei) <a href=\"https:\/\/twitter.com\/drfeifei\/status\/1987891210699379091?ref_src=twsrc%5Etfw\">November 10, 2025<\/a><\/p><\/blockquote>\n<p> <script async src=\"https:\/\/platform.twitter.com\/widgets.js\" charset=\"utf-8\"><\/script><\/p>\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>\u201cSuch systems must generate spatially coherent worlds that adhere to physical laws, process multimodal inputs\u2014from images to actions\u2014and predict the evolution of these worlds,\u201d Li explained.\u00a0<\/p>\n<\/blockquote>\n<p>According to the professor&#8217;s vision, spatial intelligence represents \u201ca frontier beyond language\u2014the ability to create interconnections.\u201d<\/p>\n<h2 class=\"wp-block-heading\"><strong>The Concept of \u201cWorld Models\u201d<\/strong><\/h2>\n<p><a href=\"https:\/\/www.nvidia.com\/en-us\/glossary\/world-models\/?ncid=pa-srch-goog-634983&#038;_bt=767867925040&#038;_bk=world+model&#038;_bm=p&#038;_bn=g&#038;_bg=189198925651&#038;gad_source=1&#038;gad_campaignid=22866024845&#038;gbraid=0AAAAAD4XAoGxv2psFYMWu5azHq9avfpNi&#038;gclid=CjwKCAiAt8bIBhBpEiwAzH1w6Q9XyITg5ILuJ3Jv-9M7B3JanEepQiH2s1jhZkS6XUS1ENKfrdJd8hoC0DkQAvD_BwE\">The concept<\/a> emerged in the early 1940s during the research of Scottish philosopher and psychologist Kenneth Craik in cognitive science.\u00a0<\/p>\n<p>The idea resurfaced in the modern AI space in 2018 following a <a href=\"https:\/\/arxiv.org\/pdf\/1803.10122\">paper<\/a> by David Ha and J\u00fcrgen Schmidhuber, suggesting that a neural network could learn and recreate a compact internal model of its environment and use it as a simulator for planning and control.<\/p>\n<p>However, solving the problem requires creating complex systems capable of storing spatial memory and modeling scenes in more than two dimensions.<\/p>\n<p>In September, Li&#8217;s company, World Labs, released a beta version of <a href=\"https:\/\/www.worldlabs.ai\/blog\/bigger-better-worlds\">Marble<\/a>\u2014an early \u201cworld model\u201d that created interactive three-dimensional environments using text or graphic prompts.<\/p>\n<figure class=\"wp-block-video\"><video controls src=\"https:\/\/forklog.com\/wp-content\/uploads\/img-8f932a11a096d457-9268366690089015.mp4\"><\/video><\/figure>\n<p>Users could navigate the generated environments without time constraints or scene loading, while the environment remained unified, unchanged, and intact.<\/p>\n<figure class=\"wp-block-video\"><video controls src=\"https:\/\/forklog.com\/wp-content\/uploads\/img-4cfd8a62f88a8e5d-9268363295743660.mp4\"><\/video><figcaption class=\"wp-element-caption\">Example of Marble in action. Source: World Labs.<\/figcaption><\/figure>\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>\u201cThe next frontier in AI development will be spatial intelligence\u2014a technology that will turn vision into reasoning, perception into action, and imagination into creativity,\u201d said Li, describing Marble as merely the first step.<\/p>\n<\/blockquote>\n<p>In October, Nvidia <a href=\"https:\/\/forklog.com\/en\/news\/nvidia-unveils-a-system-to-link-quantum-computers-to-its-ai-chips\">introduced<\/a> a system for connecting quantum computers to the company\u2019s AI chips. The technology will significantly accelerate data processing and open new opportunities for research in medicine and materials science.\u00a0<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Artificial intelligence is not yet fully capable of understanding the physical world. This remains the primary challenge for the technology.<\/p>\n","protected":false},"author":1,"featured_media":90893,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"select":"1","news_style_id":"1","cryptorium_level":"","_short_excerpt_text":"AI struggles to fully grasp the physical world, posing a major challenge.","creation_source":"","_metatest_mainpost_news_update":false,"footnotes":""},"categories":[3],"tags":[438,167],"class_list":["post-90892","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news-and-analysis","tag-artificial-intelligence","tag-research"],"aioseo_notices":[],"amp_enabled":true,"views":"196","promo_type":"1","layout_type":"1","short_excerpt":"AI struggles to fully grasp the physical world, posing a major challenge.","is_update":"","_links":{"self":[{"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/posts\/90892","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/comments?post=90892"}],"version-history":[{"count":1,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/posts\/90892\/revisions"}],"predecessor-version":[{"id":90894,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/posts\/90892\/revisions\/90894"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/media\/90893"}],"wp:attachment":[{"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/media?parent=90892"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/categories?post=90892"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/forklog.com\/en\/wp-json\/wp\/v2\/tags?post=90892"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}