Highlights:

• Living rat neurons successfully trained to perform real-time AI computations
• Structured neuron networks improved learning and reduced synchronization
• Technology shows strong potential for brain-machine interfaces and bio-AI systems

Research Overview

Researchers developed a novel bio-AI system where living rat cortical neurons were trained to perform real-time computational tasks. The study focuses on combining biological neural networks with machine learning techniques using a closed-loop reservoir computing approach. The goal is to explore whether living neurons can act as functional computing systems rather than just biological components.

System Design and Working

The system integrates living neurons with high-density microelectrode arrays and microfluidic devices. Neural signals are recorded, converted into continuous outputs, and fed back into the system as electrical stimulation in a loop of about 330 milliseconds. A real-time learning method continuously adjusts the output to match target signals, allowing the system to learn without external intervention.

Network Structuring Innovation

To improve performance, neurons were physically organized into 128 micropores connected through microchannels. This design prevented excessive synchronization, which is common in unstructured neural networks. As a result, neuron correlation dropped significantly from 0.45 to around 0.12, leading to more complex and efficient network behavior. The lattice network structure showed the best overall performance.

Capabilities and Results

The system successfully generated different waveform patterns such as sine, square, and triangular waves across multiple time intervals. It also demonstrated the ability to approximate complex chaotic systems like the Lorenz attractor. During training, the system maintained strong accuracy, achieving correlation levels above 0.8.

Limitations and Future Scope

Despite its capabilities, the system shows a performance decline after training stops, with increasing error during autonomous operation. A key limitation is the 330-millisecond feedback delay, which restricts the system’s ability to handle fast-changing signals. Future work aims to reduce latency using specialized hardware, with potential applications in brain-machine interfaces, neural prosthetics, and next-generation bio-hybrid AI systems.

Read More:

• Xiaomi’s New Electric Shaver Pro Set is Insane – 95 Days Battery & Smart Cleaning
• Inside Project Prometheus: The Secret AI Startup of Jeff Bezos, That’s Hiring Top xAI Talent
• Huawei Unveils Vision Smart Screen 6 SE: 288Hz 4K Mini LED TV Goes Official

(via)

The post Rat Brain Cells Trained to Perform AI Tasks in Groundbreaking Study appeared first on Gizmochina.