Khosla Ventures Backs Ex-Google AI Leaders' New Startup

A new AI startup, Discovery Loop, has emerged with a mission to significantly accelerate scientific experimentation. The company was founded by former leaders at Google, including Jeff Dean, the company's chief scientist, who brings a wealth of experience from innovations like TensorFlow, TPUs, and Gemini. Khosla Ventures co-led the seed round for Discovery Loop, signaling strong investor confidence in the venture. Samir Kaul, managing director at Khosla Ventures, joined the program to discuss the startup's ambitious goals and the vision behind its formation.

The Genesis of Discovery Loop

The formation of Discovery Loop is attributed to the collective expertise of its founders, who collectively spent 80 years at Google. When this group, including Jeff Dean, approached Khosla Ventures with their vision to build a company focused on engineering and research to improve science, Kaul found it an opportunity too compelling to refuse. "When guys like Jeff Dean and his team who are responsible for so many innovations like Google, TensorFlow, TPUs, Gemini, they come to you and say, look. We're coming up with a, we're gonna leave after a collective eighty years at Google to start a new company that's gonna be based in engineering and research and improving science, how do you say no?" Kaul remarked.

The full discussion can be found on Bloomberg Technology's YouTube channel.

Automating the Experimental Loop

Discovery Loop aims to automate the entire experimental process. By leveraging frontier AI models and large-scale computational infrastructure, the startup plans to rapidly propose, run, and learn from experiments. This approach is designed to drastically compress iteration time and increase the quantity and quality of scientific and engineering output. Kaul elaborated on the potential applications, stating, "AI is gonna design experiments. It's gonna test and figure out what the results of that experience are gonna be, and then it's gonna design another thousands and thousands of parallel experiments in a closed loop system to give you answers about, you know, what type of magnets should be used for fusion, types of battery materials for solid state batteries, for for electric vehicles, for drug discovery, for drug development."

A Focused Approach in a Crowded Field

While major AI labs like OpenAI, Anthropic, and Google DeepMind are also exploring scientific domains, Discovery Loop differentiates itself through its singular focus. "Jeff's sole focus with Sanjay and and his O'Day and Kaul are gonna be RSI. That's it for science and engineering. He doesn't have 15 other, masters to respond to. That's gonna be his sole focus. He's gonna clearly be able to recruit a killer team and, and stay focused on that big prize," Kaul explained.

The Valuation and Public Benefit Corporation Structure

The investment decision was reportedly swift, driven by the founders' track record and the immense potential of the AI-driven scientific discovery market. Kaul noted that the opportunity is comparable to the impact of large language models, which have already generated significant market value. Discovery Loop is structured as a public benefit corporation, a move Kaul believes will attract engineers passionate about solving significant scientific challenges. He drew a parallel to Khosla Ventures' early investment in OpenAI when it was a nonprofit, which proved to be a highly profitable venture.

Regarding the current startup funding climate, Kaul acknowledged the shifting definition of "seed stage," where companies can raise substantial capital early on. He asserted that for Discovery Loop, the "prize is big enough that it warrants the round and the valuation that Jeff and his team have have achieved." He also reflected on the changing perception of startup valuations, noting how the definition of a "unicorn" has evolved significantly since 2005.