Robots Are Having their ChatGPT Moment

Dear Artisan,

Robots are about to have their ChatGPT moment.

While most of the AI advancements of the past 24 months have been focused on software and creating better ways of retrieving and organizing information, a new release from robotics startup Covariant could bring that same methodology to the physical world.

Let’s dive in.

The One Big Thing

Moving AI From Bits to Atoms

It’s been a crazy 24 hours in AI. Famed AI Twitter personality Beff Jezos announced his latest startup, Extropic, a new way of creating computing resources using thermodynamics. The US Government commissioned a report on AI threats that turned out to be even more fear mongering than you would expect even from the USG.

But the news from Covariant trumps all of that in a landslide. Today, we have our first ever Robotics Foundation Model: RFM-1.

The Basics:

• RFM-1, a Robotics Foundation Model, has been developed to accurately interact with and manipulate the physical world, utilizing both internet data and extensive real-world robotics interactions.

• The model is trained on a diverse and multimodal dataset, collected from Covariant's deployed robotic systems in various real-world environments, to understand complex physical interactions and achieve high performance.

• RFM-1's capabilities extend to physics simulation, language understanding for intuitive robot programming, and predictive world modeling, aiming to make robots more autonomous and efficient in complex tasks.

• Despite its groundbreaking advancements, RFM-1 faces limitations such as resolution and frame rate constraints, reliance on traditional programming for orchestration, and the need for further scaling to enhance its understanding and operational capabilities.

Covariant's unveiling of RFM-1 isn't just another incremental step; it's akin to planting a flag on previously uncharted territory. It’s a shift towards enabling machines with a level of understanding and interaction with the physical world that mirrors human intuition.

The potential future applications of this innovation are what make it most compelling. By blending vast datasets from the digital and physical realms, RFM-1 is essentially being taught to 'think' and 'act' in a world filled with unpredictable variables. This isn't just about making robots smarter; it's about redefining the boundaries of machine capability.

RFM-1 is a foundation model that all robots can use in years to come. The stated goal is for this to be used by “billions” of robots in the coming years. Its multi modal approach means that the model allows for robots to understand natural language, while creating 3D models of the world that they inhabit, allowing for human-machine interactions that could have previously been impossible.

Rather than learning to program a robot in the future, engineers and even factory workers will be able to prompt their robot assistants with prompts like “move the bolt a 2.2 inches to the right and 1 inch higher.”

Now you don’t need to be a creative genius to see where this could go awry. Combining bits and atoms has proven difficult through the history of technology and this will be no different. I expect many months of testing and perhaps regulation to play out before we see commercially available robots using RFM-1.

This is, however, a new paradigm we are entering in the age of AI. You can just feel the acceleration.

The Gallery

Zuck, Snoop and Leo as Vintage Samurai

Trailer for the future of humanity

Maybe Claude 3 Is Not AGI After All

Tools

Must-have tools for every artisan to add to their toolkit:

Deep Tech

The newest and coolest in the research world that you need to know about:

Closing Thought

Work With Us!

The AI Renaissance is coming and we are building the best community of the people making it happen.

Contact us to sponsor your product or brand and reach the exact audience for your needs across our newsletter and podcast network.