Naveen Rao: 4D Computing, AI's Energy Wall & Beating Biology

The session opens by introducing Naveen Rao as co-founder and CEO of Unconventional AI, an AI chip startup, and as someone best known for building and selling two deep tech companies. The opening also includes a business claim: a business of about $20 million when the person speaking arrived had grown to "$7 or $800 million" by the time they left. The transcript does not name that business at this point.

I don't think you really understand something until you can build it. Just because something that's tried does not mean it's wrong.

The speaker calls themselves "the opposite of an AI dumer" and describes AI as "the next evolution of humanity." They also argue that true intelligence requires innovation in the hardware substrate. They call the event "sort of the anti-dumer conference" and present AI as one of the most transformational technologies humanity has created.

The speaker then gives a short personal history. They had a computer at home very early, in 1978, and learned to program as a child because it felt like a puzzle. They became an electrical engineer because they loved science fiction and wanted to make an intelligent machine. After a career building computers, they earned a PhD in neuroscience to return to the question of how to make computers intelligent.

The speaker says they founded what they describe as the first AI chip company, Nirvana Systems, in 2014. At that time AI was not part of common vocabulary, and it was hard to convince people the field mattered, let alone build hardware for it. They sold the company to Intel, a sale they now think came "way too early," and then started and ran Intel's AI group.

After leaving Intel in 2020, the speaker moved to the infrastructure for training larger models, such as today's large language models. Their team turned GPUs into a platform that could scale and that other people could use easily. After ChatGPT arrived in 2022, the speaker says the company was "kind of the best game in town" for people building their own models, and it grew quickly. It joined Databricks in 2023. According to the speaker, that business now accounts for "a quarter of the total revenue of Databricks today."