We are pleased to announce our investment in Archetype. Find out why we have invested in the company and learn more about their foundation models and use-case specific applications for real-world sensor data.
The Inevitable Shift: The Evolution of Multimodal Foundation Models to Bridge the Sensory Gap
Large language models (LLMs) have become a cornerstone of AI, demonstrating remarkable capabilities in text, image and even video generation. However, their reliance solely on a few input data types presents limitations. Humans are inherently multimodal. We process and react to a combination of text, speech, visuals, touch, and other sensory cues to understand the world around us and convey meaning and intent. To fully explore the potential of Physical AI, LLMs must evolve — and the future lies in multimodality and the ability of large models to truly understand the world in the ‘physical,’ sensory sense.
Archetype AI’s foundation model, Newton is built to move beyond the digital realm into the physical world. Built as a developer platform, it will empower the developer community within the industrial segment and beyond to benefit from its multimodal capabilities to solve their real-world problems and build applications.
We at Hitachi Ventures, are excited to announce our participation in the seed round of Archetype AI led by Venrock with participation from Amazon Industrial Innovation Fund, Buckley Ventures, Plug & Play Venture Group and Strange Ventures.
Why Archetype AI?
Archetype AI is taking a revolutionary approach to foundation models by building a Multimodal Foundation Model that truly understands the physical world. This means it can interpret and translate complex data from sensors like cameras, radars, and thermal sensors into human-friendly formats like text, visuals, and even code. The Archetype AI team has devised a novel approach enabling users to query real-time multimodal sensor data to perceive, understand and predict physical world behavior, including human behavior.
Sensor data is very different from web data and is complex to understand. It comes in various modalities, is multidimensional, and needs to be evaluated and understood in a time series. Archetype AI’s multimodal approach fuses the sensor modalities together for various object detection, classification, and prediction capabilities and ingests into a LBM (Large Behavioral Model). The LBM takes the latent representation and gains understanding based on context to determine human behavior and predict intent. The model incorporates knowledge of physics and temporal awareness to model out the dynamics of the physical world.
With this novel approach, we expect to see multiple use cases being unlocked across industries including Robotics, Automotive, Consumer Devices, Healthcare, Insurance and Retail. Archetype AI has already signed a strategic partnership with Infineon to accelerate the future development of sensor-based chips with advanced AI functionalities. We are also excited about the potential prospect of such a novel technology for a large industrial conglomerate like Hitachi!
We believe we have the right team to bring this to fruition at Archetype AI. The core team spent eight years together at Google ATAP (Advanced Technology & Projects) division building a deep expertise in sensor decoding and encoding and using AI to model sensor data to detect and determine human intent. We believe they have the right skills and experiences to build out the future of multimodal foundation models.
The Road to the Future
While multimodal foundation models are still under development, research is progressing rapidly, and we believe Archetype AI represents a paradigm shift in AI by bridging the gap between the digital and physical world — paving the way for a future where machines interact with the world in a more nuanced, versatile, and human-like way.
You can sign up for early access to Archetype AI’s developer platform here.