Two startups want to replace how AI learns: one just raised $180M, another is seeking up to $1B

Two startups—Flapping Airplanes and Core Automation—are raising hundreds of millions to research new AI architectures that require far less training data.

Flapping Airplanes raised $180 million from GV, Sequoia, and Index. The startup wants to build AI systems that learn more like humans do, without "without ingesting half the internet." The company estimates that humans are 100,000 to 1,000,000 times more data-efficient than current models.

That would require changes at a foundational level: new loss functions, possibly replacing gradient descent altogether and some "weird, new ideas." The team is focused on research for now, with potential applications in robotics, retail, and science down the line.

Former OpenAI researcher starts competing lab

Jerry Tworek, a former senior researcher at OpenAI, founded Core Automation and is seeking $500 million to $1 billion in funding, The Information reports. His goal is to build models that need 100 times less data and can learn continuously from experience, something he considers essential for general artificial intelligence.

Core Automation's long-term plans include AI agents that would develop products ranging from industrial automation to "self-replicating factories," bio-machines, and ultimately even planetary terraforming.

Tworek says he left OpenAI because this type of fundamental research wasn't a priority there anymore. Current architectures can still improve, but they have limits, and that creates a need for new breakthroughs. Other startups founded by OpenAI alumni include Thinking Machine Labs and Safe Super Intelligence.