The Analog Iterative Machine (AIM) Offers Unparalleled Speed
Image Source: MicrosoftA team of researchers at Microsoft has successfully built the first 8-variable optical computer, named AIM that is Analog Iterative Machine, which harnesses different intensities of light to compute and store information simultaneously. The breakthrough computer has demonstrated remarkable potential in solving complex optimization problems with superior speed and capacity compared to traditional binary systems.
Led by Christos Gkantsidis, one of the principal researchers on the project, the team initially aimed to accelerate machine learning using AIM. However, their focus shifted when they discovered the computer's unparalleled ability to address optimization challenges, which underpin critical sectors such as finance, healthcare, logistics, and manufacturing.
Recognizing the significance of this advancement, Barclays Bank PLC has entered into a one-year research agreement with Microsoft to explore the application of AIM in real-world scenarios. The partnership aims to tackle the problem of settling batches of transactions at clearing houses, a task that is often hampered by the sheer scale of data involved.
Lee Braine, managing director and distinguished engineer at Barclays, explained that evaluating all possible options for transaction settlement would take an impractical amount of time using conventional binary computers.
The AIM team successfully solved a "toy version" of the settlement problem posed by Barclays with 100% accuracy, highlighting the potential of optical computing in this domain. By contrast, previous attempts using different technologies only achieved a 50% success rate. Excited about the prospects, Braine and the Microsoft researchers are now working on a larger-scale version of the problem, intending to test it on an upgraded version of AIM later this summer.
The development of AIM and its optical computing capabilities has gained significance due to the plateauing of Moore's Law, which predicted the doubling of transistors in integrated circuits every two years. As traditional computing approaches reach their limits,
Alternative technologies like optical computing are being explored. Hitesh Ballani, one of the Microsoft researchers involved in the project, emphasized the urgency to develop such alternatives, noting that optics presented the most challenging obstacle to overcome.
Also read:- What is Microsoft AIM and How it's works ?
Microsoft's Cambridge-based research lab has previously made strides in optical storage, devising a system that stores vast amounts of data within glass. Now, with AIM, the team is taking optical computing to new heights. Operating at the speed of light, advanced iterations of AIM are expected to outperform binary computers by a factor of about a hundred. Furthermore, AIM's unique feature of integrating computation and storage at the same location sets it apart from traditional computers, which require separate memory and computing units.
The success of AIM also relies on groundbreaking algorithms developed by the research team. Francesca Parmigiani, a primary researcher responsible for building the optical computer, stressed the critical importance of co-designing hardware and algorithms to achieve optimal performance. The team has created a novel type of algorithm called QUMO (quadratic unconstrained mixed optimization), specifically tailored for AIM, making the computer unique in the world.
As the researchers continue to test AIM and refine their algorithms, they are actively seeking additional real-world problem cases from industry experts and academics. A service utilizing an AIM simulator has been launched, solving large-scale optimization problems using graphics processing units (GPUs). Among the priority challenges is the transaction settlement problem proposed by Barclays, which, if successfully addressed, could pave the way for leveraging optical computing in various other banking applications, including fraud detection.
Lee Braine expressed his excitement about being at the forefront of this groundbreaking research, highlighting the opportunity to push boundaries and contribute to innovative change. Ant Rowstron, the team lead and Microsoft Distinguished Engineer, commended the collaborative effort of the multidisciplinary group in inventing this new type of computer tool. He emphasized the importance of supporting researchers in exploring new frontiers and their belief that AIM will offer a promising solution to pressing problem spaces.
With the rapid advancements made by the AIM team, the future of optical computing appears brighter than ever, offering a potential paradigm shift in solving complex optimization problems across various industries.