Former Openai and Deepmind researchers are growing a whopping $300 million in seeds to automate science

The regular lab came out of stealth on Tuesday with a $300 million war chest as a seed round. It is supported by the technology industry.

The regular lab was founded by Ekin Dogus Cubuk and Liam Fedus. Cubuk led the materials and chemistry teams for Google Brain and Deepmind. For example, one of his projects was an AI tool called Gnome. The tool discovered over 2 million new crystals in 2023. This is a material that can be used one day to drive a new generation of technology, researchers say.

Fedus was Openai’s former Vice President of Research and was one of the researchers who helped create ChatGpt. He also led the team that created the first trill-parameter neural network.

Similarly, the small team is filled with researchers who have worked on other major AI and materials science projects, from building an agent operator for Openai to working on Microsoft’s Pathergen, the LLM Materials Science Discovery AI.

The goal of the regular lab is nothing more than automating scientific discoveries and generating AI scientists, the company says. This means that the robot will run physics experiments, collect data, iterate, retry, and build a lab that will learn and improve.

The lab’s first goal is to invent a new superconductor that performs better than existing superconducting materials and probably requires less energy. But funded startups are also hoping to find other new material.

Another goal is to collect all the physical world data that AI scientists generate as they search for new things and mix and manipulate different forces and raw materials.

“To date, advances in scientific AI have been progressing from models trained on the Internet,” and LLMS has “used out” the internet as a source of consumption, the company says in its introductory blog post. “[A]Periodically, we have built and autonomous laboratories with AI scientists and they are run. ”

The hope is that not only will the lab invent next-generation material, but will also generate invaluable fresh data that AI models can consume to continue their evolution.

This may be one of the most impressive group of researchers to frame startups for this purpose, but that’s not the only one working on AI scientists. AI as a tool for automating chemical discovery has been a topic of academic research since at least 2023. This is the pursuit of small startups like Tetsuan Scientific and nonprofits like Future House and Toront University of Toronto’s Acceleration Consortium.