The Happy Hour with Architects hosts prominent researchers from academia and industry to discuss different topics related to computer architecture. We hope to create a sense of community during this pandemic by providing a platform to discuss and debate research trends and directions in computer architecture and systems. Each episode hosts two guests who share a common interest. The guests engage in an informal, free-format conversation and express their excitement and vision for the field. In this episode, I am super excited to discuss the research opportunities and challenges at the intersection of neuroscience and computer architecture with Tim Sherwood and Abhishek Bhattacharjee. Our discussion spanned two main directions. First, designing new computation models taking inspiration from how the brain works, and second, designing systems to understand our brains and treat neurological disorders.
Tune in to this episode of Happy Hour with Architects to learn about the research directions at the intersection of neuroscience and computer architecture.
Designing Neuro-Inspired Hardware
Data encoding in the brain is time-based; the temporal relationship between the arrivals of spikes represents useful information. Taking inspiration from this representation, Tim Sherwood and his team asked a simple research question: what happens if we try to do computation with values encoded only as times?
This research led to “race logic”, where information is represented as a timing delay and the time it takes to complete a task provides the final result. And the end result is a simpler and energy-efficient design and can be particularly useful in the embedded space to classify images or process sensor data. Abhishek mentions that another particularly interesting new use case of temporal logic can be designing embedded systems for interacting with the biological neurons. This leads to our second direction of designing systems to understand our brains.
Designing Hardware to Interface with Neurons
Having a fast and flexible interface to the brain has a board implication; it can help us to understand our brains and accelerate the treatments for neurological disorders (e.g., epilepsy, Parkinson’s disease, anxiety, and schizophrenia). However, in order to enable that future, not only we need to be targeting millions of neurons in the brain, we also need to make sure that we can stimulate brain regions for different treatments in real-time with extremely low power and thermal limit.
To this end, Abhishek and his team, in collaboration with Rajit Manohar, took on a novel and ambitious goal: how to build low-power, flexible, and fast hardware for brain-computer interface devices that can monitor electromagnetic signals from the neurons and can process those signals for the treatment of various types of neurological disorder? His work led to a general-purpose architecture for implantable brain-computer interfaces.
Tune in to this episode of Happy Hour with Architects to learn more about this amazing research topic. Tim and Abhishek discuss their vision of where the field is heading and how future architects can take on the responsibility of integrating neuroscience and computer architecture to enable a new future.
About the Author
Samira Khan is an assistant professor at the University of Virginia and the host of the “Happy Hour with Architects”. The goal of her research group, ShiftLab, is to introduce a paradigm shift in the current computing system by fundamentally rethinking our current processor-centric computing model.
