\n<\/p>\n
While the rest of the AI industry races to label its work as \u201cAGI\u201d or \u201csuperintelligence,\u201d Alexandre LeBrun, the CEO of Yann LeCun\u2019s<\/a> world model<\/a> startup, AMI Labs,<\/a> avoids the terms altogether. \u00a0Lebrun said in an interview with TechCrunch that the company doesn\u2019t use terms like \u201cAGI\u201d or \u201csuperintelligence\u201d at all.<\/p>\n \u201cWe never used the word AGI. And I just noticed that nobody is using it anymore; they switched to superintelligence,\u201d he said. \u201cNext time we\u2019ll switch to something else.\u201d He isn\u2019t sold on the new label either. \u201cThere\u2019s no good definition. What is superintelligence? I don\u2019t know. It\u2019s not a very useful word.\u201d<\/p>\n It\u2019s a pointed stance from a founder sitting at the center of AI\u2019s newest race. <\/p>\n TechCrunch talked to LeBrun while he was in Seoul last week for The International Conference on Machine Learning, where he was scouting for local industrial partners, global companies and researchers. AMI Labs is still pre-product, but it\u2019s already courting robotics, manufacturing and electronics players. A world model, which incorporates physics to predict and work with the real world, needs to prove itself outside the lab, LeBrun explained.<\/p>\n One area where world models are expected to have a large impact is robotics. For now, robots are just running fixed routines, \u201ccompletely static\u201d, and AI remains \u201creally dumb in the physical world,\u201d LeBrun said.<\/p>\n Even when AI can merely make robots \u201caware of the context\u201d that would mark \u201ca very big difference for the world.\u201d Such context-aware AI would have been useful, for example, in preventing a robot that was dancing and doing kung fu at a public event<\/a> from approaching and kicking a child. \u201cThe hardware is very advanced; progress in hardware in the last few months is incredible, but there\u2019s no brain.\u201d<\/p>\n A large language model (LLM) predicts the next word or text, and a world model predicts the next state. Nudge a glass off the table, and you already know it will tip and spill; that\u2019s the intuition a world model is meant to capture: predicting the next state of the world, LeBrun explained.<\/p>\n He isn\u2019t claiming world models are better than LLMs, which are \u201ccomplementary, not replaceable\u201d when it comes to AI systems that understand the physical world, LeBrun said. Drawing a parallel to the human brain\u2019s distinct language and reasoning functions, he added that LLMs will remain the most efficient tools for processing language while world models will provide context and real-world understanding.<\/p>\n Almost every industry that \u201ctouches the real world\u201d could eventually make use of robotics based on world models, LeBrun said, arguing that physical environments remain where LLMs are weakest.<\/p>\n A factory robot repeating the same motion works well enough today, he said. The challenge begins when \u201cyou take your robot outside into a more open environment, in your household, or in the street\u201d, where it must understand its surroundings and operate safely. \u201cRobots are not safe right now,\u201d he said. \u201cThere\u2019s no solution for that today.\u201d<\/p>\n Healthcare offers a more personal example for LeBrun, whose previous company was Nabla, an AI health startup. He likened today\u2019s AI systems to a doctor trained only on textbooks and without a residency. LLMs may be useful in medicine, he said, but they cover \u201conly 1% of healthcare.\u201d The rest depends on real-world experience. \u00a0<\/p>\n But a world model, LeBrun said, can\u2019t be built inside a lab. To train on reality, AMI needs real environments and close partners, according to the CEO. \u201cWe need access to the real world,\u201d and it\u2019s \u201ceasier for us to do that with partners.\u201d That is part of what pulls him toward Asia, where the robots, chips, and factories actually are.<\/p>\n LeBrun won\u2019t spell out a full Asia strategy yet. \u201cIt\u2019s too early,\u201d he said. But the pull toward South Korea comes down to two things. First, Korea has advanced industries in robotics, semiconductors and manufacturing; the hardware-heavy sectors that the first wave of AI barely touched.<\/p>\n The second attraction is speed. LeBrun pointed to Korea\u2019s national plan to pour money into AI and its track record as an early adopter. \u00a0\u201cKorea was the fastest adopter of the internet 25 years ago,\u201d he said. It\u2019s that combination, a deep industrial base plus a willingness to embrace AI fast, that he calls \u201cunique,\u201d and the reason \u201cwe want to be here from day one.\u201d<\/p>\n \u201cI\u2019ve been telling Alex and the team to come to Korea,\u201d JP Lee, the CEO of SBVA and one of AMI\u2019s backers in Asia, told TechCrunch. <\/p>\n The government has done \u201ca tremendous job\u201d funding local sovereign LLM models, Lee said, and those already work \u201cwell enough\u201d for general-purpose tasks, but he\u2019s pushing for Korea to keep investing in physical AI, too. He points to Seoul\u2019s June plan to mobilize some $880 billion<\/a> for chips, AI data centers and physical AI, as one of its three declared pillars. \u201cThey should coexist.\u201d<\/p>\n Korea\u2019s value to foreign firms, Lee argued, isn\u2019t only in hardware. Local developers are quick to adopt and adapt new tools, a pattern that has produced homegrown internet players like Naver and Kakao. <\/p>\n For all the star power and the billion-dollar check, AMI has nothing to sell yet. The startup, co-founded by Turing Award winner Yann LeCun<\/a> after he left Meta, raised $1.03 billion in March<\/a> at a $3.5 billion pre-money valuation. There\u2019s no product yet, and no timeline he\u2019ll commit to. \u201cWe\u2019ll make a surprise when we\u2019re ready,\u201d LeBrun said.<\/p>\n<\/div>\n When you purchase through links in our articles, we may earn a small commission<\/a>. This doesn\u2019t affect our editorial independence.<\/em><\/p>\n