Why AMI Labs’ Alexandre LeBrun won’t call his AI ‘AGI’ or ‘superintelligence’

In a landscape where the term 'AGI' and 'superintelligence' are rapidly becoming buzzwords, Alexandre LeBrun, the CEO of AMI Labs, is taking a different approach. In a recent interview with TechCrunch, LeBrun clarified that his company avoids these labels entirely. "We never used the word AGI, and I've noticed that no one is using it anymore; they’ve switched to superintelligence," he remarked, expressing skepticism about the usefulness of these terms. LeBrun's remarks come at a crucial time as the AI industry pushes towards new capabilities. Speaking from Seoul during The International Conference on Machine Learning, he highlighted AMI Labs’ ongoing search for partnerships in robotics, manufacturing, and electronics. Although the company is still in its pre-product phase, it aims to revolutionize how robots interact with the real world. The concept of a world model—an AI framework that incorporates physics to predict real-world scenarios—could be transformative, particularly in robotics. Currently, many robots operate on fixed routines, which LeBrun describes as “completely static.” He emphasized that even the ability to make robots aware of their context would represent a significant leap forward. For instance, such an advancement could prevent a robot, which is programmed to perform choreographed movements, from accidentally harming a child at a public event. Despite advancements in hardware, LeBrun pointed out a critical gap: the lack of an intelligent 'brain' for these robots. He explained that while large language models (LLMs) predict text, world models aim to forecast the next state of the world. "If you nudge a glass off the table, you know it will tip over and spill—that’s the intuitive understanding a world model is designed to capture," he elaborated. LeBrun does not claim that world models surpass LLMs; rather, he sees them as complementary. He drew a parallel between the distinct functions of language and reasoning in the human brain, asserting that LLMs are superior for language processing while world models enhance real-world understanding. This technology has the potential to impact numerous industries that rely on physical environments, an area where LLMs often struggle. The challenge intensifies when robots are taken out of controlled environments and into more dynamic settings like homes or streets, where they must navigate safely. LeBrun voiced concerns about current robot safety, stating, "Robots are not safe right now; there’s no solution for that today." LeBrun also shared insights from his previous venture in healthcare, comparing today’s AI systems to a doctor who has only studied textbooks without practical experience. He noted that while LLMs can assist in medical tasks, they only cover a fraction of healthcare needs. A world model, he believes, cannot be developed in isolation and requires real-world training environments and partnerships. His interest in Asia stems from the region's advancements in robotics and manufacturing, as well as its history of quickly adopting new technologies. LeBrun pointed out Korea's robust industrial base and its ambitious national plan to invest in AI, which he considers a unique opportunity for AMI Labs. JP Lee, CEO of SBVA and an AMI backer, echoed this sentiment, emphasizing the local government's commitment to funding AI initiatives. He suggested that Korea’s strengths in hardware and rapid adaptation of new tools create a conducive environment for foreign firms. Despite raising $1.03 billion in March at a valuation of $3.5 billion, AMI Labs has yet to launch any products, with LeBrun indicating that the company will reveal its innovations when they are fully prepared.