As large language models (LLMs) dominate the artificial intelligence market, a prominent voice has argued they still possess clear limitations in understanding the physical world. Simultaneously, new research showing that a human-like reasoning structure spontaneously formed inside an LLM has reignited the debate over artificial general intelligence (AGI).<\/p>\n
Pascal Fung, co-founder and Chief Research & Innovation Officer (CRIO) of AMI Labs, delivered a keynote at the International Conference on Machine Learning (ICML) 2026 at COEX in Seoul on July 7. \u201cLLMs only understand the world indirectly through text written by humans,\u201d Fung said. \u201cFor AI agents that operate in the real world, we need a world model that directly understands the physical environment.\u201d<\/p>\n
Fung used a soccer match as an analogy. A player instantly reads the spatial relationships between the ball and opponents, the causal links between physical actions and outcomes, and even grasps team objectives, communication with teammates, and emotions. \u201cA true world model reads both the physical and mental worlds together like this,\u201d he explained.<\/p>\n
AMI Labs is a startup founded late last year by Yann LeCun, a professor at New York University, former chief AI scientist at Meta, and a globally recognized AI authority. The company drew industry attention by raising $1 billion (~1.5 trillion won) in seed funding this year alone. LeCun has consistently argued that LLMs alone face clear limits in extending AI into the real world.<\/p>\n
Fung pointed to the learning objective as the fundamental problem with LLMs. \u201cThe direction of LLMs aims for fluent conversation and linguistic expression, not understanding the world,\u201d he said. \u201cThey are ill-suited to grasping the physical causality of the real world.\u201d He added, \u201cBecause LLMs operate by generating the next token through massive computation, they are much slower and more costly than humans even when trying to understand the world.\u201d<\/p>\n
The hallucination problem of LLMs was also cited as an obstacle to real-world expansion. \u201cHallucinations occurring at the text level may be harmless, but if a hallucination occurs in a robot, it could lead to a collision,\u201d Fung warned.<\/p>\n
Indeed, even top-tier models like OpenAI\u2019s GPT and Anthropic\u2019s Claude fall far short of humans in physical world understanding. On DeepPhy, a benchmark evaluating models\u2019 physical reasoning performance, humans achieved a 64.7% accuracy rate, while the highest score among AI models was just 41.2%.<\/p>\n
Fung presented perception, prediction, planning, and memory as essential elements for AI agents operating in the real world. He also noted that current Vision-Language-Action (VLA) models mostly rely on imitation learning and become vulnerable when they deviate from the training data distribution, stressing that \u201ca world model not tied to specific actions is needed.\u201d<\/p>\n
AMI Labs is researching world models based on technologies like JEPA (Joint Embedding Predictive Architecture), which LeCun has studied since his time at Meta. Unlike LLMs that generate the next pixel, JEPA focuses on understanding the world by predicting the next situation. This path differs from video-generation AI extensions like Nvidia\u2019s Cosmos or Google DeepMind\u2019s Genie.<\/p>\n
\u201cPhysical AI is an area where many things remain unsolved,\u201d Fung said. \u201cWe need robotic agents that can learn in a zero-shot manner even in environments with far less data.\u201d<\/p>\n
Anthropic Finds Traces of AGI Inside Claude<\/p>\n
At the very moment LeCun was emphasizing the limits of LLMs, Anthropic announced research findings that it had discovered evidence approaching AGI inside its own Claude model.<\/p>\n
According to a paper published by Anthropic, \u201cGlobal Workspace in Language Models,\u201d a special region called \u2018J-space\u2019 exists inside Claude. This region controls Claude\u2019s thinking and reasoning; without J-space, Claude reportedly drops to a level where it can barely answer kindergarten-level questions.<\/p>\n
Notably, J-space is not something Anthropic intentionally designed but a structure that Claude spontaneously formed over its long training process. The Anthropic research team was ultimately unable to determine exactly how J-space emerged.<\/p>\n
Drawing an analogy to human brain activity, most tasks\u2014breathing, posture adjustment, etc.\u2014are processed unconsciously, while conscious thought occurs intensively in specific regions. Anthropic explained that a similar division appears inside Claude. Most tasks are processed automatically, but when complex reasoning is required, information concentrates in J-space, a \u2018concept space.\u2019<\/p>\n
Using an observation tool called the \u2018Jacobian lens\u2019 (J-lens) developed by the researchers to peer into J-space\u2019s workings, it was revealed that when Claude answers a question, it first recalls key concepts in J-space and then develops its reasoning based on them. For example, when asked, \u201cHow many legs does an animal that spins webs have?\u201d Claude first activated \u2018spider\u2019 in J-space, then answered \u2018eight\u2019 based on that.<\/p>\n
Even more striking is that artificially altering a concept in J-space completely changes Claude\u2019s answer. When researchers changed \u2018spider\u2019 in J-space to \u2018ant,\u2019 Claude immediately answered \u2018six.\u2019 Changing \u2018France\u2019 to \u2018China\u2019 caused it to answer questions about France with China\u2019s capital and currency. This is evidence that J-space is not merely a record of results but directly participates in the subsequent reasoning process.<\/p>\n
Anthropic mapped J-space\u2019s characteristics onto the properties of \u2018consciously accessible information\u2019 discussed in human cognitive science: the content of J-space can be reported, it is actively activated on demand, a single concept is flexibly utilized across multiple tasks, and it operates selectively.<\/p>\n
This discovery provides a significant turning point in the AGI debate. In the past, a model\u2019s AGI status was judged by external performance such as solving math problems, coding ability, or passing exams. However, the J-space case demonstrates the possibility that models can internally and spontaneously form human-like reasoning structures.<\/p>\n
LeCun\u2019s AGI Skepticism<\/p>\n
Ahead of this, LeCun took to his X (formerly Twitter) account on July 4 and bluntly stated, \u201cThe \u2018G\u2019 in AGI is nonsense.\u201d Since 2023, he has consistently argued that \u201cChatGPT, Claude, and Gemini are not the path to human-level intelligence.\u201d<\/p>\n
LeCun\u2019s logic is clear. Exam questions are verbalized, discretized tasks\u2014precisely the area where LLMs excel. True intelligence must include abilities stemming from common sense, such as perception, physical intuition, and causal understanding.<\/p>\n
In an interview, he offered the following example: \u201cWhat happens if you stand a pen upright and let go? Even a young child knows the pen will fall. But no one can predict which direction it will fall. An LLM will generate a plausible prediction based on statistical patterns, but because it is performing statistical completion rather than reasoning about physical reality, it will almost certainly be wrong.\u201d<\/p>\n
LeCun argues that AI agents must observe the world, learn how the world changes, and then internally formulate predictions and plans before acting. LLMs have learned vast linguistic knowledge, but this is merely an interface for interaction, not an understanding of the world itself.<\/p>\n
He also opposes the Silicon Valley belief that simply scaling up model size will cause AGI to emerge spontaneously. His reasoning is that language data alone cannot fully capture the real world, and pure autoregressive prediction is not an efficient method of reasoning.<\/p>\n
However, what LeCun opposes is less the concept of AGI itself than the Silicon Valley tendency to overhype current AI systems as imminent AGI and emphasize doomsday risks. He warns that this attitude inflates excessive expectations for AI and could ultimately cause the AI bubble to completely collapse.<\/p>\n
Alongside CRIO Fung, AMI Labs Chief Scientific Officer (CSO) Xining Xie and others were also reported to have visited South Korea for ICML 2026. During the conference, AMI Labs plans to hold a joint event with SBVA (formerly SoftBank Ventures) to share its research direction with global researchers.<\/p>\n","protected":false},"excerpt":{"rendered":"As large language models (LLMs) dominate the artificial intelligence market, a prominent voice has argued they still possess…\n","protected":false},"author":2,"featured_media":98683,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[4],"tags":[6744,42830,53,3013,182,51306,51455,1642,51454,33463,6968],"class_list":["post-98682","post","type-post","status-publish","format-standard","has-post-thumbnail","category-agi","tag-agi","tag-ami-labs","tag-anthropic","tag-artificial-general-intelligence","tag-claude","tag-j-space","tag-jepa","tag-large-language-models","tag-pascal-fung","tag-world-model","tag-yann-lecun"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/98682","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/comments?post=98682"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/98682\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media\/98683"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media?parent=98682"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/categories?post=98682"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/tags?post=98682"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}