Flapping Airplanes raises $180 million to rethink AI data efficiency

In an interview last week, the co-founders of Flapping Airplanes—brothers Ben and Asher Spector, alongside Aidan Smith—outlined their vision for the newly launched artificial intelligence research lab. The company has raised $180 million in seed funding to challenge the industry’s focus on scaling existing transformer models. While established labs like OpenAI and DeepMind have spent heavily to scale their Large Language Models (LLMs), Flapping Airplanes is focused on finding less data-hungry ways to train AI. According to Ben Spector, current frontier models are trained on the sum totality of human knowledge, whereas humans can make do with significantly less. The startup aims to address this gap by prioritizing data efficiency over sheer scale.

The lab’s approach is guided by a philosophy reflected in its name. Rather than trying to replicate the human brain entirely, the founders seek to build alternative architectures that operate under different constraints. Aidan Smith, who previously worked at Neuralink, a brain-computer interface company, noted that the brain is not the ceiling for AI capabilities, but rather the floor. The team plans to draw inspiration from biological systems without being bound by them, contrasting their work with the massive scale of current models developed over the last five to 10 years. Ben Spector, a co-founder of the lab, explained the analogy: “Think of the current systems as big, Boeing 787s. We’re not trying to build birds. That’s a step too far. We’re trying to build some kind of a flapping airplane.”

This architectural shift could unlock new capabilities. Asher Spector stated that it is possible that as models are trained on less data, they are forced to have deeper understanding. Such models would probably be a million times more data efficient, making them far easier to integrate into data-constrained environments. To achieve this, the founders are building a research team composed of young, creative talent, including individuals still in college or high school. The co-founders believe that younger researchers bring fresh perspectives because they have not been constrained by the established paradigms of thousands of academic papers. Smith emphasized that the lab wants to try radically different approaches, even if those attempts occasionally yield worse results than the current paradigm.

Why it matters

Flapping Airplanes represents a shift in the AI research landscape, moving away from the “scaling laws” paradigm toward fundamental architectural changes. These alternative approaches could make AI models significantly more efficient and capable in data-constrained environments.