Last week, we made waves at TEDAI San Francisco — on Day 1, we premiered our feature video, "A Phenomenological AI Foundation Model for Physical Signals," and on Day 2, our co-founder, CEO, and CTO, Ivan Poupyrev, took the stage for a panel discussion on the future of embodied AI. Read on to get an insight into how Archetype AI is addressing some of the biggest real world challenges with Newton.
AI Learning Physics from Sensor Data
Our feature video debuted with a bold vision: addressing some of the world’s toughest challenges – across healthcare, energy, transportation, manufacturing, and infrastructure – with AI solutions. It introduced Newton, demonstrating how this foundation model can learn principles of physics directly from real world sensor data, making it capable of tackling a vast range of complex, physical world challenges – such as optimizing energy use, improving infrastructure, and enhancing manufacturing processes.
The video showcased a milestone in our journey toward developing a physical AI foundation model. Traditional methods involve “teaching” AI about the physical world through structured data and preset laws, like physics equations. But our approach takes a new direction: enabling AI to learn through direct observations gathered from diverse sensors. Without relying on predefined data or pre-existing knowledge, the model decodes sensor input, understands it, and even predicts future physical behavior, allowing it to solve problems across an unprecedented range of applications. This adaptability brings AI closer to an augmented intelligence– one that complements, rather than replicates, human understanding and opens up countless possibilities for real world problem-solving.
Transformative Power of Embodied AI’s
In the panel discussion, alongside Field AI’s Sebastien De Halleux, NVIDIA’s Amit Goel, Principal Venture Partners’ Songyee Yoon and Bloomberg’s Shirin Ghaffary, Ivan shared his perspective on how embodied AI is poised to transform the physical world and the work underway at Archetype AI to bridge the gap between AI and the real world. He dove deep into the philosophy behind Archetype AI and the potential of embodied AI to operate seamlessly in real world settings. He described the development of a “foundation model for the physical world” – an AI capable of making sense of the diverse and complex sensor data that define our environments. From spectrograms to radar, our model is designed to interact with this data intuitively, creating insights that drive physical-world solutions.
During the panel, Ivan shared that one of the critical challenges for embodied AI is ensuring that robots or other other types of equipment and machinery can make decisions independently in unstructured environments, without requiring step-by-step human guidance. This challenge led partially to the development of our foundational model, Newton, which autonomously discovers and analyzes patterns within its surroundings. He likened this development to creating a “ChatGPT for the physical world,” with Newton offering autonomous insights that humans may not have even considered.
Bridging Physical and Digital Worlds
Together, our video premiere and panel discussion highlighted the potential for physical AI to adapt and innovate in real time, providing a glimpse into a future where technology doesn’t just support human capabilities but actively expands our understanding of the physical world.
Stay tuned as we continue to share advancements that expand AI’s impact across industries, making once-unsolvable challenges attainable.
