We bring AI to the physical world through
discovery-based autonomous learning
Our Mission
AI Is Bound To Digital Space
The current AI revolution is based over models that can only process and produce text. Hence, it is bound to digital space, where text is the main way of interaction.
But Real Impact Lies In The Physical World
Only 15% of the economy is based in the digital world. The other 85% is out of reach of the AI revolution because it would need a way to interact physically with the world.
And Current Models Can Only Get You So Far
All the approaches to translate the current AI revolution to the physical world have failed. Their main limitation is always the huge amount of learning required to deal with the complexity of the real world.
Our Approach
1
Instead Of Learning From Canned Samples...
Training datasets are good for reproducing scenarios, but they are not the way intelligent beings learn. This is a limitation of current approaches due to the learning mechanism they use.
2
... Let AI Discover The World Like A Baby Does...
Our technology allows the AI to discover its own body and its surrounding world in real time, without needing datasets or pre-training. This allows the AI to build its own world model like an intelligent being does.
3
... For The Better Of Humanity!
Our technology allows the development of robots that can supply the labor shortages present in some critical industries, and/or replace workers in dangerous works where humans are at risk of death.
Discover our Technology:
Embodied Discovery Models (EDMs)
Brain Inspired Neuron Architecture
Anchored in latest scientific discoveries, our patent-pending Cortical Computational Unit (CCU) is our implementation of a neuronal column as defined by the Thousand Brains Theory of Jeff Hawkins.
Embodied, Child-Like Discovery Mechanism
Based on real-world experimentation, our learning mechanism builds its abstractions and reasoning over the physics of the real world. This process is performed in real time, without costly pre-training phases.
Local Learning Rule for Neurons
The key element allowing us to get where no other AI has being able to get is our novel, patent-pending learning mechanism. It is based on a local learning rule that updates each neuron independently, while ensuring stability and convergence.
Embodied Discovery Models
This implies that in current AI models more neurons are a trade-off between cost and intelligence, while in EDMs more neurons increase intelligence without trade-offs.
An AI For General Purpose Robots
The next step in the Physical AI revolution
Adaptable
To the most unpredictable environments
Efficient
Learning on the fly, without costly pre-trainings
Secure
With guardrails on place to avoid risks
An AI For Autonomous Driving
A new revolution in the logistic chain
Quick Reaction
To unforeseen circumstances
Cost Effective
No need for huge data collection efforts to train the AI
Safe
Reliable driving without human intervention
Meet our founders
The dedicated professionals driving our success
Piotr Ciochoń
Chief Executive Officer
Piotr brings his experience as co-founder of two deep-tech startups, and his expertise as Physics Ph.D., to drive the business model and operations of our company.
Alfredo Ibias
Chief Scientist Officer
Alfredo brings his experience as outstanding AI researcher to develop the AI of the future. He also brings his expertise as lead scientist, as ex-META employee, and as author of 20 AI papers, to drive the science of our company.
Guillem Ramírez
Chief Technology Officer
Guillem brings his expertise at the Barcelona Supercomputing Centre (BSC), and as an expert in scalable AI platform development, to steer the deployment strategy and infrastructure of our company.