Speaker: Oshiorenoya E. Agabi, PhD
About the Seminar:
It’s evident that we have reached the limit of silicon technology. Over the last 5 years, processor performance and transistor counts have largely stalled while costs are shooting up! So, what are the alternatives? We have neuromorphic chips, quantum computers, 3D transistors with “fins”, and graphene-based processors on the table. However, a key ingredient for the success of silicon is scale. How do each of these systems scale? How do we characterize the power performance curves in each of these devices? Interestingly, these devices are based on similar technologies — silicon, or a variation thereof. As such, it’s prudent to investigate alternative technologies — especially systems based on soft matter that is biology. Biology is indeed the most advanced technology on the planet.
We have also observed that massive shifts in communication and computation have occurred with a change in substrate. Koniku believes that applying neurons as a material just like silicon will usher in the next change, and help us meet the impending computational and communication challenges.
This talk will expound upon Koniku’s progress in building a processor using real biological neurons, and the company’s immediate (3-5 years) and long term goals to increase the neuron density on our chips. Lastly, there will be a discussion about the types of applications, which may be possible with such devices.
About the Presenter:
Oshiorenoya Agabi (Founder & CEO, Koniku Inc.), has over 10+ years of experience in neuroelectronic interfacing within both industry and academia. As a strategic program lead at a robotics startup in Switzerland (Neuronics AG, acquired in 2008), Osh implemented learning algorithms for pick-and-place robots which work alongside factory workers. The team achieved significant success in developing robots that did not learn object categorization autonomously.
Osh also led a cross disciplinary industry team to develop an in vitro reflex arc for modeling implantable neural chips. These chips were to interface with the peripheral nervous system. The project was completed with a successful design of an implantable chip lasting 30-40 years. While leading work in the applications of liquid state machines to biological neurons on chip, this was completed for the first time in 2004 at the ETH Zurich (Eidgenössische Technische Hochschule). Here, he developed applied criteria for computability of biological neurons in vitro.
As a visiting scholar and PhD candidate at the Imperial College in London, Osh built 2 photon microscopes for studying synaptic transmission in the mouse. He also worked on developing super resolved automated systems for stimulating optogenetic neurons with addressable light systems. The intent was to understand the mapping of visual signals with thalamocortical slices in vitro — that is, “the various stages of visual information processing in the brain.”