Flapping Airplanes and the promise of research-driven AI

On Wednesday, a pioneering AI lab named Flapping Airplanes officially opened its doors, backed by an impressive $180 million in seed funding from prominent investors including Google Ventures, Sequoia, and Index. The founding team boasts a wealth of expertise, and their mission to discover a more efficient method for training large models is particularly noteworthy. What sets Flapping Airplanes apart is its departure from the industry's longstanding focus on scaling, which has predominantly centered around amassing vast amounts of data and computational power. According to Sequoia partner David Cahn, this lab is among the first to challenge the scaling paradigm, which advocates for dedicating extensive societal resources to enhance today's large language models (LLMs) in the pursuit of artificial general intelligence (AGI). In contrast, Flapping Airplanes adopts a research-centric approach, positing that we may be just a few significant breakthroughs away from achieving AGI. This perspective emphasizes the importance of long-term research endeavors, particularly those that may take five to ten years to yield results. While the compute-first strategy prioritizes immediate gains and massive server expansions, the research-first model encourages a broader exploration of possibilities, even if many of those efforts have a low chance of success. As the tech landscape leans heavily towards compute-centric advancements, the emergence of Flapping Airplanes signals a refreshing alternative approach, focusing on foundational research that could ultimately reshape the future of AI.