Anthropic Lets Claude Run the Office Vending Machine
AI safety startup Anthropic created 'Project Vend' to let its AI system Claudius manage snack dispensing.
Anthropic is doing something weird with its AI. The company behind Claude has launched an internal experiment called Project Vend, putting a version of its AI system—nicknamed Claudius—in charge of the office vending machine.
It sounds like a joke, but it's not. The project is part of Anthropic's broader mission to understand what's actually happening inside their AI's brain.
Key researcher Chris Olah and other executives are leading efforts to examine Claude's neurons—yes, the artificial kind—to figure out how the system actually thinks. It's AI interpretability research with a quirky real-world test case.
The New Yorker profile reveals a company obsessed with cracking open the black box of large language models. Letting an AI manage snack distribution? That's just Anthropic being Anthropic.