Key facts
- AMI Labs CEO Alexandre LeBrun stated the company avoids calling its AI 'AGI' or 'superintelligence'.
- LeBrun believes current AI is 'really dumb in the physical world' and robots are 'not safe'.
- AMI Labs is developing 'world models' that predict real-world states, complementing LLMs.
- The company is seeking partners in robotics, manufacturing, and electronics, with a focus on Asia.
- AMI Labs raised $1.03 billion in March at a $3.5 billion pre-money valuation.
Alexandre LeBrun, CEO of AI startup AMI Labs, is deliberately sidestepping the industry trend of labeling artificial intelligence as 'AGI' or 'superintelligence.' In an interview with TechCrunch, LeBrun stated that his company does not use these terms, finding them ill-defined and unhelpful. He emphasized that AMI Labs is focused on developing 'world models' that incorporate physics to predict and interact with the real world, aiming to address the current limitations of AI in physical environments.
LeBrun explained that while Large Language Models (LLMs) excel at predicting text, world models are designed to predict the next state of the physical world. He sees these two types of AI as complementary, with LLMs handling language and world models providing crucial real-world context and understanding. This distinction is particularly relevant for advancing robotics, where current systems are described as 'static' and 'really dumb' in physical settings.
AMI Labs, which is still pre-product, is actively seeking partnerships in robotics, manufacturing, and electronics. LeBrun highlighted Asia, particularly South Korea, as a strategic focus due to its advanced industrial base in these hardware-heavy sectors and its history of rapid technology adoption. He noted that Korea's government is investing heavily in AI, including physical AI, which aligns with AMI Labs' goals.
The startup, co-founded by Turing Award winner Yann LeCun, has already secured significant funding, raising $1.03 billion in March at a $3.5 billion pre-money valuation, despite not yet having a product on the market.