What makes Microsoft’s light-based computer 100 times more efficient

Microsoft has developed a new type of computer called an analog optical computer (AOC) that uses light for calculations instead of traditional electronic switches. This approach offers the potential to boost energy efficiency for AI by up to 100 times compared to current digital systems.

The AOC operates by using micro-LEDs and camera sensors to manipulate beams of light and perform calculations in a continuous feedback loop. This allows the machine to solve problems by iterating until it reaches a steady state or final solution.

How does Microsoft’s computer use light to calculate?

Unlike digital computers that rely on billions of tiny switches flipping on and off, Microsoft's system uses the intensity and voltage of light beams to add and multiply numbers. The calculations are done in analog form, enabling faster and more efficient problem solving.

The AOC repeatedly computes a problem, refining its solution with each cycle until a final answer is found. This method takes advantage of physical properties of light to accelerate complex computations important for AI tasks.

Did you know?
The concept of analog optical computing dates back to the 1960s, long before digital computers became mainstream.

What role do micro-LEDs and camera sensors play?

The computer consists of arrays of micro-LEDs to emit light and camera sensors that detect variations in light intensity. These components work together to perform mathematical operations such as addition and multiplication.

Using commercially available micro-LEDs and sensors from smartphone technology helped the researchers build a scalable and affordable prototype with existing supply chains. This combination enables the light-based computations to be practical and energy efficient.

Why is analog optical computing more energy efficient?

Traditional digital computers convert analog signals to digital ones during calculations, which consumes significant energy and imposes speed limits. The AOC avoids these conversions by operating entirely in analog.

This saves energy and reduces bottlenecks, allowing the light-based system to achieve an estimated 100-fold improvement in energy efficiency over digital GPUs for specific AI and optimization tasks.

ALSO READ | What Happened in the Salesloft GitHub Account Breach Exposing Data

What tasks can the AOC perform today?

The prototype AOC has demonstrated the ability to perform simple machine learning tasks like image classification with comparable accuracy to digital computers. It has also solved real-world optimization problems related to financial transactions and MRI image reconstruction.

These early successes show potential for the AOC to improve processing speed and energy consumption in applications requiring complex reasoning and optimization.

What is the future potential of Microsoft’s AOC?

The current prototype has limited scale, but future versions could add millions or billions of micro-LEDs, enabling powerful computations with much larger data sets. Researchers are also developing a 'digital twin' to simulate more complex problems that the physical AOC cannot yet tackle.

Microsoft aims for the AOC to become a significant part of future computing, supporting the scaling of AI while dramatically reducing energy costs and environmental impact.