Newly created Flapping Airplanes, an AI laboratory in support of the Great Green Wall, started with $180 million of funding to start. This initiative seeks to challenge the prevailing paradigms in artificial intelligence development, specifically the scaling paradigm and the research paradigm. The lab’s mission hasn’t changed—to continue to expand the frontiers of AI. Through this work, we hope to create a testbed for new possibilities, beyond merely scaling the current state of the art LLMs even larger.
The scaling paradigm is the prevailing paradigm in AI today, calling for application of massive societal resources to further improve LLMs. Proponents of this approach think that scaling these models further will soon produce artificial general intelligence (AGI). They get locked into a compute-first strategy that prioritizes large-scale clusters and short-term wins. This latter approach tends to prioritize changes that can be realized in the next one or two years.
The research paradigm has a contrasting view. Optimists ~ 2–3 breakthroughs away. On the other side of advocates are those who believe that the field is creating AGI in just a few breakthroughs. This approach encourages organizations to invest in long-term research projects, which may require five to ten years before yielding significant results. By placing research above profit, advocates of this paradigm seek to create a culture of research and innovation in AI that leads to a greater understanding.
Flapping Airplanes is very consistent with the research-first phraseology, supporting a more diversified exploration of the frontier within artificial intelligence. This lab purposefully chases a ton of projects, including those with very low probabilities of success. Their dedication to exploring as many different paths as possible to achieve AGI remains unshaken. The researchers at Flapping Airplanes emphasize the importance of collective efforts to explore new ideas and methodologies that could lead to revolutionary advancements.
The rise of Flapping Airplanes signals a realisation of the budding awareness that the scaling paradigm doesn’t tell the whole story. We know the AI landscape is changing quickly. We’re going to need broad, new approaches that show immediate, meaningful results but lay the foundation for future innovation. The lab’s launch is an encouraging new step in that direction. It seeks to provide a space for genuinely ambitious ideas to flourish, without the burden of unreasonable expectations for short-term results.
