Young artificial intelligence startup Flapping Airplanes has impressed Silicon Valley with a strong founding team and a goal to discover methods for training large models using smaller amounts of data. On Wednesday, the enterprise stated that it secured a $1 billion Series A funding round, co-led by Index and Lux, with a reported valuation of $6 billion.
This funding arrives only nine months after the business debuted with a $180 million seed round backed by GV, Sequoia, and Index. That initial seed round valued the firm at $1.5 billion, meaning the valuation has quadrupled.
Additional participants in this latest round featured past supporters Sequoia, GV, and Kleiner Perkins. Valor, Atreides, Kleiner Perkins, A*, and Nvidia likewise hold investments in the enterprise.
Aidan Smith and brothers Asher and Ben Spector established Flapping Airplanes. Smith is a previous founder and Thiel fellow who was employed at neural implant firm Neuralink during his college years. Asher Spector earned a statistics PhD at Stanford, whereas Ben Spector served as a computer science PhD student at Stanford and started Prod during his time at MIT.
Prod operates as an elusive, five-year-old student-operated startup incubator that has transformed into one of the most powerful talent incubators in Silicon Valley, according to Sequoia partner and Flapping Planes investor David Cahn in a publication earlier this year outlining his investment rationale. Combined, the creators are constructing the young person's AGI lab, as Cahn characterized it.
The creators characterize their project as the world's most complete megakernel framework. It can execute entire training runs, handle inference, operate video games, and execute all three concurrently within a single kernel. A kernel functions as a program operating on a GPU, and a megakernel combines numerous tasks into a single program.
The creators state the framework can overlap memory-bound inference with compute-bound training, meaning it can execute an operation that spends most of its duration awaiting data retrieval alongside one that spends most of its duration performing mathematical calculations, ensuring the chip rarely remains idle. The underlying concept is that artificial intelligence has barely touched the limits of GPU capabilities, and they maintain it can build intelligent systems without ingesting half the internet.
The objective consists of systems requiring lower expenses for training and operation. A model that is a million times more data efficient is probably a million times easier to put into the economy, Asher Spector stated to TechCrunch soon after the enterprise debuted.
They currently possess a new $1 billion pool to recruit personnel and try to transform that objective into a reality.
