Yann LeCun departs Meta to launch AMI startup focused on physical‑world AI

Yann LeCun, Meta’s Chief AI Scientist and head of the Fundamental AI Research (FAIR) lab, is leaving the company after 12 years to start his own AI venture. He says the new startup will continue the Advanced Machine Intelligence (AMI) program he has pursued at FAIR, NYU and beyond, and will partner with Meta on some initiatives.

LeCun described the startup’s goal as building systems that better understand the physical world, have persistent memory, can reason and plan complex action sequences — capabilities he believes won’t come from simply scaling large language models. He will remain at Meta through the end of the year while the new company ramps up.

Context and why this matters

• Research direction: LeCun has been vocal about the limits of LLM scaling and is pushing for architectures that combine perception, memory and planning.
• Company shifts: Meta has recently reorganized its AI efforts, invested heavily in other AI partners and appointed new senior AI leadership — factors that likely influenced LeCun’s decision.
• Partnership model: The startup will partner with Meta, suggesting a collaborative approach rather than an adversarial split — and potential access to resources or data under agreed terms.

Key questions to watch

• How the AMI startup will fund and scale research compared with large incumbents (OpenAI, Google, Anthropic).
• Whether Meta’s partnership gives the new company strategic advantages or limits its independence.
• How LeCun’s emphasis on physical understanding and planning will shape architectures and benchmarks for next‑generation AI.

LeCun thanked Meta leadership for support of FAIR and the AMI program, noting the new entity aims to maximize the broad impact of this research across many economic sectors. As the AI field diversifies beyond purely language‑centered models, this move could accelerate alternative approaches to general intelligence.

For original reporting and further details, see the coverage: Engadget.

Discussion: Do you think pursuing AMI outside a big tech lab will accelerate progress on embodied, memory‑aware AI — or will independence slow access to the compute and data needed? What should LeCun’s team prioritize first?