#United States #San Francisco #AI Robotics #HYPRLABS #HYPRDRIVE

HYPRLABS Emerges with a New Paradigm in AI Robotics

In an exciting development for the realm of artificial intelligence and robotics, HYPRLABS Inc., an innovative company with bases in San Francisco and Paris, has officially launched its groundbreaking architecture known as HYPRDRIVE™. This new technology is poised to redefine how robots learn and operate, emphasizing a direct interaction with their environments to enhance their capabilities.

A Shift from Traditional Robotics

What sets HYPRDRIVE™ apart from existing systems is its rejection of traditional reliance on pre-coded rules, extensive training data, and high-definition maps. Instead, the architecture utilizes a novel approach called Run-time Learning (RTL), which allows robots to learn at a higher velocity by engaging with their surroundings in real-time. By doing this, HYPRDRIVE™ enables robots to convert their experiences directly into intelligence, minimizing the loss of information that often occurs when using simulated environments.

Emphasizing Learning Velocity

Tim Kentley Klay, the Co-Founder and CEO of HYPRLABS, highlights the significance of Learning Velocity in their approach. According to him, this metric represents how quickly a robot can transform exposure into valuable insights. The shift towards continual, live learning for robots distinguishes HYPRLABS from its competitors and establishes a new frontier in robotic autonomy.

The Three-Phase Learning Pipeline

HYPRDRIVE™ operates on a sophisticated three-phase continual learning pipeline:

1. Foundational Learning: Robots are initially operated by humans, establishing a base of real-world behaviors without relying on fragile pre-coded instructions or approximation models.
2. Hybrid Learning: This phase involves human supervision where feedback can be provided to the AI in real-time. This way, the AI can adjust and learn from immediate corrections, allowing for a dimension of reinforcement learning from human feedback.
3. Continuous Learning: Signals from other robots in the fleet are gathered in the cloud to refine models continually. This allows quick adjustments to robots with minimal data transfer, ensuring that learning is both rapid and efficient.

Efficiency and Performance

Backed by data gathered from real-world applications, HYPRDRIVE™ has recently demonstrated its impressive capabilities by successfully maneuvering through a demanding route in downtown San Francisco. The test vehicle, featuring a minimal hardware stack of just five vision cameras and a single NVIDIA Orin AGX, operated on a mere 33 watts of power—equivalent to that required to charge a smartphone. This efficiency underlines the potential for deploying robust robotic systems even on a limited power budget.

The Emergence of Domain-General AI

With a focus on creating domain-general autonomy, HYPRLABS is moving beyond the limitations of programmed behaviors in robotics. According to Werner Duvaud, Co-Founder and Director of AI at HYPR, transitioning to an adaptable, learning-based model enables robots to thrive in various environments and tasks without pre-existing conditions.

The journey for HYPRLABS is just beginning, with plans to introduce its first line of novel robots powered by HYPRDRIVE™ by 2026. Their future products are anticipated to outperform existing technologies across key markets, setting a new standard for intelligent robotic movement.

For more fascinating insights into HYPRDRIVE™ and how it shapes the future of robotics, visit HYPR's official website.

About HYPRLABS

Founded to revolutionize AI and robotics, HYPRLABS is dedicated to creating autonomous systems that learn efficiently, directly from their environments. With roots in two vibrant cities, San Francisco and Paris, they are harnessing the potential of live learning to forge ahead in the technology landscape.

About Tim Kentley Klay

Tim Kentley Klay is not only the Co-Founder of HYPRLABS but also the original architect of Zoox, recognized for its pioneering contributions to the autonomous vehicle industry. His impressive portfolio includes 125 patents and over 100 design accolades.

As HYPRLABS unveils its innovative technology, the possibilities for the future of robotics become even more intriguing, illuminating a path toward smarter, more autonomous machines.