Complex landscape for gradient descent
Cloud Systems Futures

AIM (Analog Iterative Machine)

Header image by Javier Ideami @ losslandscape.com (opens in new tab)

chart

In Project AIM, analog iterative machine, we are building an optical machine to solve hard optimization problems at the speed of light — a space where state-of-the-art silicon solutions and even quantum computers fall short.  

Our machine has an asynchronous data flow architecture and is built using commodity opto-electronic technologies that are low-cost and scalable. It offers speed-ups in excess of 100x compared to existing state-of-the-art digital approaches. It also offers a new abstraction that, unlike existing optimization hardware, is better suited to practical problems. This combination of a massive speed-up for solving real problems accurately at scale and at low cost captures our optical machine’s potential to power optimization in the post-Moore Law’s era.   

Project AIM is part of the broader Cloud Systems Futures project, whereby we aim to innovate across the cloud stack by co-designing the cloud’s software and hardware infrastructure. Our cross-disciplinary team thus spans the entire stack, with experience in systems and networking, optics (system, sub-system and device-level), and hardware. 

Update (2023-06-08): We are pleased to announce an online service that will allow participants to experiment with the AIM algorithm. For now, the service provides a performance GPU-based implementation of the AIM algorithm, which will allow the users to experiment with converting problems to the QUMO abstraction and understand the benefits and shortcomings of the AIM algorithm. We plan to provide access to the hardware in the near future — stay tuned.
We currently have capacity for a limited number of users. Please let us know if you are interested to participate by writing to us at project-aim-contact@microsoft.com.