Technology

Fusion AI with Digital Brains leveraging state of the art modern AI

Unique Technology Assets

inait is building on 20-years of research by 1’100 scientists and engineers on finding the recipe to digitally replicate the brain’s biological design, backed by $300M and led by inait’s founder Henry Markram. inait has developed the proprietary operating and learning mechanisms needed to teach digital brains. inait has also built technologies for a wide range of industries to connect digital brain technologies with modern AI, perfecting fusion AI.

Our core pillars

Digital Brains as the neural network architecture already evolved for generalizable intelligence

A Neural Code as the foundation of a Digital Brain Operating System to program encoding, processing and decoding computations in digital brains

A Causal Learning Rule as a singular learning rule to teach digital brains to acquire cognitive skills

Fusion AI as a holistic AI ecosystem with digital brains at its center and leveraging modern state-of-the-art AI, such as LLMs, CNNs, and GNNs

Digital Brains

The Open Brain Institute is making the five step recipe for building digital brains available to spark the Age of Digital Brain Building across all species. inait has developed the technology to convert these digital brains into neural network architectures that incorporate as much biology as computational power is needed for a particular use case.

Demonstration 1

inait has deciphered the neural code and developed a Digital Brain Operating System to provide programmatic control over this “language of the brain,” managing how neurons process, encode, and decode information. The system allows spiking neural networks (SNNs) with architectures as complex as the brain’s, to be programmed to function the way the real brain does.

Processing data with digital brains

Input
An image as a list of pixel values

Processing
Circuit activity in response to stimulation

Read-out
Measure emergent geometric structures and turn into bits

Interpretation
Final output based on this set of bits

A Digital Brain Video Codec

Original Movie

Neural Activity

Neural Code

Decoded Movie

The neural code has holographic-like properties enabling unprecedented data compression, robustness to missing or corrupted information, and resistant to adversarial attacks.

Demonstration 2

inait has taken the already winning learning rule that Markram discovered for SNNs in a completely new direction. By operating entirely different from backpropagation used in modern AI to learn patterns and correlations in static datasets, inait’s AI can learn cause and effect to link actions to consequences, allowing learning from experience and embodiment, much like any brain. inait’s causal learning was tested on benchmarks in the OpenAI gym, solving them with a fraction of the size of network and time to learn compared to deep reinforcement learning.

Physics AI benchmarks

iAM Architecture

inait has developed the connectors to couple digital brains to modern, state of the art AI–leveraging them for vision, hearing, speech and knowledge to create a wholistic AI that perceives, adapts intelligently to real, virtual and digital environments and can communicate directly with customers to get instructions and explain what it is “seeing, thinking and doing”. We call these systems Intelligent Action Models (iAMs).

The digital brain delivers advanced cognition and interactive adaptability, while state-of-the-art pattern detection AI provides precise sensory pre-processing. Additionally, large language models add a communication layer, enabling the system to express its actions and decisions in text and even command code as needed.