Advanced Machine Intelligence (AMI), a fledgling firm co-established by computer science innovator and former Meta AI Chief Executive Officer Yann Lecoun, revealed on Tuesday that it has procured $1.03 billion to engineer “world models” — artificial intelligence capable of grasping and engaging with the tangible realm.
The monetary backing for AMI, situated in Paris, represented the largest initial funding tranche for a European startup to date and ranks among the most significant investments in AI enterprises across the area generally, as per Crunchbase. The funding effort was spearheaded by Bezos Expeditions, Cathay Innovation, Greycroft, Hiro Capital, and HV Capital, valuing AMI at $3.5 billion subsequent to the investment.
AMI stands apart from well-known generative AI startups by focusing on the creation of “world models” — artificial intelligence with the ability to interact and assimilate knowledge derived from three-dimensional reality.
“I anticipate that ‘world models’ will emerge as the next sought-after term,” AMI Labs Chief Executive Alexandre Lebrun conveyed to TechCrunch following the investment. “Within half a year, every entity will be labeling itself a ‘world model’ to garner financial support.”
Co-founder Yann Lecoun is recognized as a leading figure in the Large Language Models (LLM) methodology within AI. In 2018, he was bestowed with the distinguished AM Turing Award in recognition of his contributions to neural networks and learning algorithms.
However, AMI's fresh venture rests on the premise that the two-dimensional methodology employed by LLMs is inherently constrained. AMI executives maintain that for AI to achieve genuine utility, it must comprehend and engage with three-dimensional reality.
“Generative frameworks educated via supervised learning emulate intelligence; they do not genuinely perceive the world. Token prediction, although potent, performs optimally for discrete and low-dimensional tasks such as information retrieval, summarization, coding, and mathematics,” AMI CEO Alexandre Lebrun articulated on LinkedIn. “However, factories, hospitals, and occupations in open settings necessitate AI that understands reality. And reality is not tokenized: it is continuous, fluctuating, and high-dimensional. Despite their capability, I do not believe generative frameworks will culminate in authentic understanding of the world.”
AMI's initial collaborative undertaking is with the medical AI startup Nabla, also headed by Alexander Lebrun.
Source: crunchbase