AI Breakthrough: New Chip Promises to Revolutionize Many Fields

A chip can see movement, analyze and remember that movement, just like the human brain, but in just a few microseconds (Illustration: ST).

This research from scientists at RMIT University (Australia) promises to revolutionize the robotics, autonomous vehicle and intelligent systems industries, helping them interact with the environment faster, more efficiently and significantly more energy-efficiently than current technology.

The chip is designed based on a brain-simulating structure, specifically the way neurons integrate and transmit signals.

The key material used is molybdenum disulfide, an ultra-thin (only one atom thick) material with superior optoelectronic properties.

It allows the chip to simulate the "integrate and fire" behavior of biological neurons, generating an electrical response when exposed to light.

The big difference from traditional digital systems is that this chip does not need to process the entire image.

Instead, it focuses only on detecting important visual changes, such as movement in the field of view, through an edge detection mechanism.

This approach enables superior processing speed and energy efficiency, allowing information to be processed without the need for an external computer.

The potential of this technology is particularly great in fields that require instantaneous response speeds.

Professor Sumeet Walia, Director of the COMAS Centre at RMIT, shared: "This chip could in the future be applied in autonomous vehicles or assistant robots, allowing them to respond almost immediately to dangers or human interactions."

By minimizing the amount of data that needs to be processed, this innovation contributes significantly to reducing latency, a key factor for operational safety and efficiency.

The team is working to expand from single-cell prototypes to more complex pixel arrays, targeting specific applications in embedded systems.

The project, funded by the Australian Research Council, aims to optimise the technology for a variety of visual tasks and minimise energy consumption.

This development marks a major step toward lightweight, responsive embedded AI systems that are designed not to replace the human brain but to be deeply inspired by how it works.

Detailed research is published in the journal Advanced Materials Technologies.

Embedded artificial intelligence systems are AI systems that are directly integrated into devices or hardware systems, allowing them to autonomously perform tasks of recognition, prediction, or decision making based on data collected from the surrounding environment.

Source: https://dantri.com.vn/cong-nghe/dot-pha-ai-con-chip-moi-hua-hen-cach-mang-hoa-nhieu-linh-vuc-20250519152458177.htm