2023-06-28 11:05:00
The Microsoft Research Lab in Cambridge, UK, has built a new computer that can solve optimization problems at the speed of light: the “Analog Iterative Machine“ (AIM).
Unlike the typical binary computer, it uses different intensities of light to calculate in the same place where the information is stored. In the future, such computers might exceed the speed of conventional computers by a factor of a hundred. The team is now testing the computer on optimization tasks from the banking sector. If the experiments are successful, AIM might pave the way for the use of optical computing in other specific mathematical problems.
AIM: Hundreds of times faster than conventional computers
Originally, the researchers hoped to use AIM as a tool to accelerate machine learning. However, they soon realized the real potential of the new device: to far exceed the speed and capacity of conventional binary computers, which only calculate with zeros and ones, when solving optimization problems. The new development is reacting to two current trends: the declining growth in computing capacity per dollar in digital chips, i.e. the foreseeable abolition of the decades-old Moore’s Lawas well as overcoming the limitations of previous special purpose computers designed to solve optimization problems.
IM is a novelty analog optical computer that uses photons and electrons to process continuous value data. Its architecture is completely different from today’s digital computers, which use transistors to process binary data. The Research Lab team believes that advanced versions of AIM will outperform binary computers by a factor of 100 when it comes to solving these types of problems. It’s not a general-purpose computer, say the research team, but it’s very useful for speeding up certain mathematical use cases — linear and nonlinear algebra, for example — that are currently bottlenecked.
Optimization problems: New ways for efficient calculations
Optimization problems arise in many of the most important areas of society, for example in banking and finance, healthcare, logistics and manufacturing. The promising possibilities of the new AIM computer have therefore led to a year-long research agreement with Barclays Bank. Together, the teams are investigating whether AIM can be used to solve a real problem: transactions at the clearing houses that most banks use. Their number is in the hundreds of thousands every day. And as with most optimization problems, it is scale that often overwhelms binary computers.
The AIM team already had what they call “Game Variant” of the transaction optimization problem posed by Barclays, and the optical computer solved the challenges with 100% accuracy every time. A previous attempt at research that used a different technology to address the same problem only hit the mark regarding fifty percent of the time. The Microsoft team is now working with Barclays to design a larger version of the problem with more data and variables. They hope to test them with an updated version of the AIM optical computer later this summer. “We figured that once we built it, we’d figure out how to evolve it.”explains Ant Rowstron, Microsoft Distinguished Engineer and leader of the AIM team. “And now we have an important problem area that needs urgent attention and where our computer really shines.”
www.microsoft.com/de-de
1687961172
#Microsoft #computer #solves #problems #speed #light