Please welcome P-1 AI, the company building engineering AGI for the physical world
Meet P-1 AI, the latest addition to Lerer Hippeau’s portfolio.
HQ: San Francisco
Founder: Paul Eremenko, Aleksa Gordić, Adam Nagel, Sandeep Neema, Susmit Jha
The problem: At present, the majority of AI development has been focused on the digital and not the physical world, and that’s principally because it’s so challenging to develop AI for the physical world. One of the main obstacles to creating AI models that reason on physical product domains is the scarcity of training data. The number of unique designs tends to be orders of magnitude smaller than is needed to train a large neural network, and what designs do exist tend to lack cohesive multi-physics models and are closely guarded as proprietary. So where is the AI that will ultimately help us design and build the physical systems of the future?
P-1 AI’s solution: P-1 AI is building an AI architecture that can generalize and scale to an engineering superintelligence for physical system design. The company has created the link between AI and the physical world through Archie, an AI agent aimed at the cognitive automation of tasks that human engineers perform today when designing physical products. Leveraging multi-physics and spatial reasoning, Archie is capable of distilling key design drivers from requirements, developing product concepts and derivatives, performing first-order design trades, and selecting and utilizing the right engineering tools for detailed design. Importantly, Archie is not designed to replace, improve, or compete with any existing engineering tools. Instead, its focus is on the automation of cognition at the level, initially, of a junior engineer.
Origin story: In creating P-1 AI, the co-founders brought together a strong background in model-based engineering for physical systems, which Paul, Sandeep, and Adam pioneered at scale initially at DARPA and then in several industrial applications including at Airbus and United Technologies Corporation (now RTX), together with a world-class expertise in deep learning, large multimodal models, and neurosymbolic AI, with Aleksa hailing from Google DeepMind and Susmit from leading neurosymbolic AI efforts at SRI.
Why we’re betting on it: What P-1 AI is doing with Archie, and what it aims to do over the long term, cannot be done with the typically thin wrappers around existing LLM technology. It requires a new perspective and fundamental breakthroughs in both data and models. Archie’s first deployment will begin later this year to help data center cooling systems. P-1 AI plans to rapidly expand its presence to other product domains spanning the built and mobile world, including industrial systems, building systems, automotive and heavy machinery, and aerospace and defense. We believe that the team, with their deep understanding of AI and physics, have identified a new solution to the problem of engineering AGI that has the potential to transform the way we’ll construct the future.
Our take: Obviously, the world is changing quickly as a result of this moment in AI. But the majority of AI companies have been using AI to replace outmoded functions and augment the status quo. P-1 AI is different: their mission is visionary. Some of what they’re going for might sound like science fiction, but the same could have been said about our present reality just twenty years ago. We couldn’t be more excited to partner with the P-1 AI team as they work to turn the possible into the actual.
Further reading: “Exclusive: What If AI could design a jet engine — or even a starship? Google DeepMind and Airbus veterans just raised $23M with an eye on that future” via Fortune
