Microsoft

Microsoft Developed Analogue Light Computers

Microsoft has reported success in developing a new kind of computer that doesn’t use transistors in semiconductor chips but uses light to move through multiple modular arrays. The development of the system took three years.

The new system is called the Analog Iterative Machine (AIM). AIM not only works technically differently than you are used to from conventional computers, but the binary operations are also not found here. Instead, light is passed through multiple layers of mutable materials – LED panels – that act as “modular arrays”.

For the analogue calculations, the employees at Microsoft Research in London only used fairly everyday components. Among other things, MicroLED displays and photo sensors from smartphone cameras are installed. These can work with the information space of almost the entire light spectrum, which offers enormous possibilities.

However, the AIM is not intended to replace conventional computers. Because such a device is not particularly well suited to coping with everyday tasks of the widest range. However, analog computing offers enormous advantages in certain areas, for which Microsoft is now already marketing the first systems of this type to customers.

Unbeatable in some operations

Analog computers show their strengths, for example, when solving optimization problems. For example, when it comes to searching through and evaluating gigantic amounts of data with constantly changing information according to certain criteria, classic binary computers are confronted with exponentially growing numbers of vector-matrix multiplications that simply overload a normal CPU. An analog system like the AIM, on the other hand, can easily handle this.

The first customer who will use the system is the finance company Barclays. The AIM wants to use this to track massive money transactions in stock transactions. Clearing houses, in particular, need to keep track of when mass transactions are taking place, where each actor is subject to different constraints such as different national regulations and available balances. Entire data centers are employed to manage this. In a one-year test run, the AIM should now show how much more efficiently and cheaply it can solve this problem.