Jeff Bezos backs $500 million bet on a brain-inspired AI startup

A New York-based startup called Flourish has raised $500 million to pursue one of AI’s most ambitious goals: building an artificial intelligence system that works like the human brain. Steven Levy reports for WIRED that Jeff Bezos led the early funding after reading a two-page pitch document, eventually contributing tens of millions of dollars and telling the founders he would have given more if asked.

The company was co-founded by Thomas Reardon, a neuroscientist and serial entrepreneur, and Rob Williams, a former Amazon executive who oversaw products including Alexa. Their core argument is straightforward: current large language models (LLMs) are inefficient and cannot learn after training. A single AI chip consumes more than 30 times the energy the human brain uses to process information. Flourish wants to close that gap dramatically.

What the team is actually building

The startup’s stated goal is a synthetic AI system that runs on 50 watts or less, learns continuously, and requires far less training data than today’s models. Reardon points out that a human baby learns language from a few hundred thousand spoken examples. LLMs, by contrast, require billions.

To find answers, Flourish has hired around two dozen neuroscientists and AI researchers. They plan to run original laboratory experiments using high-end equipment including electron microscopes, focusing on structures in the brain called cortical columns. These are described by one team member as “the canonical computational unit” of the brain.

Greg Wayne, a senior researcher at DeepMind who leads Google’s Project Astra, has joined as a senior adviser while keeping his existing role. Joshua Vogelstein, a co-founder, recently co-authored research showing that a fruit fly’s neural network is ten times more efficient than the transformer architecture that underpins most modern AI.

Flourish is also developing near-term products, including a memory system inspired by the hippocampus and a model designed for pocket devices. The company is in talks with a major chip manufacturer to embed its model in hardware.

Not everyone is certain the approach will succeed. Berkeley computer scientist Ben Recht, an adviser to the company, told WIRED: “I’m not convinced that it’s going to work. But if it does, it would be amazing.”