LLMs are not enough to solve the hardest problems in the world. Unlike today's AI, human experts can form strategies under uncertainty towards a goal. With incomplete information, humans take action to learn more about their environment until they land on a solution with high certainty.
In 1965, a discipline called expert systems was born in Stanford University. Its goal was to emulate the decision making ability of human experts by creating decision trees for computers. This approach was brittle because the decision trees had to be hardcoded, and the world forgot about it.
But we didn't. Over 60 years later, we formed a vision of a modern-day expert system: a computer that could form decision trees on the fly, grounded in deep domain knowledge, to reach a goal. To realize this vision required combining other forgotten AI disciplines: knowledge graphs to represent the complete space of possible reasoning paths, and reinforcement learning to learn how to dynamically navigate that space in pursuit of a goal.
We've invented a novel reasoning system that can emulate expertise in any domain. We've demonstrated that our approach outperforms the strongest LLMs on various problems under uncertainty.
Our first commercial application is in healthcare, where our technology has unlocked AI diagnostic consultations that are safe, accurate, and interpretable. Jiva, our AI resident that has clinician-level intelligence, is the path to universal and instant primary care for humanity.
We are a small, highly-coordinated team of AI researchers, engineers, designers, and doctors who think from first principles and are unafraid to challenge the status quo and invent the future. We are as much artists and philosophers as we are scientists. If our mission and DNA resonate with you and you would like to see a live demo of Jiva, please reach out to shravan@atmanlabs.ai. We are always open to meeting exceptional people who are interested in either joining our team or supporting our mission.
