🧠 Food for thought

Implications, context, and why it matters.

AMI’s $1.03 billion seed round backs a costly push for ‘world models’

• SBVA joined AMI Labs’ $1.03 billion seed round, which valued the company at a $3.5 billion pre-money price even though it has no product yet 1.
• The round supports an approach that differs from companies such as OpenAI, since founder Yann LeCun says scaling large language models will not reach human-level intelligence 2.
• AMI Labs plans to build ‘world models’ and has talked about Joint-Embedding Predictive Architecture (JEPA), a way to train AI by predicting and representing how the world changes rather than learning only from text 1.
• LeCun left Meta to start AMI because he expects the best uses to come from selling the technology to enterprises, which did not match Meta’s consumer-focused business 2.

Industry money suggests patience, while some checks may target returns

• Industrial limited partners such as Doosan and Coupang joined the round, alongside backers that include Samsung and Toyota Ventures 1.
• These firms are making early wagers on AI that can understand physical settings, which robotics and automation will need 3.
• The funding setup leans toward long research timelines since AMI Labs does not expect near-term revenue and says commercial uses could take years 1.
• A Doosan official said there are no current plans for joint projects, which suggests the stake is mainly financial rather than linked to near-term collaboration 3.