Thank you for the invitation to speak to you all today.

So before I get started, I just want to very quickly acknowledge the research group that

I co-lead here at the University of Tennessee, Tenn Lab.

So we cover all aspects of neuromorphic computing from materials and devices up through algorithms

and applications across our different faculty.

And we have many amazing students and postdocs who are working on the research topic as well,

as well as collaborators across the country.

So if you're interested in learning more about what I talk about today, you can find pretty

much all of our information on our website, which I've listed here, which you can also

reach out to me.

I'd love to answer more questions about it.

Okay, I always like to start with motivating why you should care about novel computer architectures

to begin with.

I think this was a more controversial thing to talk about several years ago, but I think

the community is starting to embrace more and more exotic hardware implementations.

There's a couple of things that started happening now about 15 years ago, 20 years ago, that

really started to revolutionize the way that we think about what we should be building

our computers out of and how we should be implementing them in terms of their architectures.

So one, of course, is the living end of Moore's law, but really the end of Dénard scaling

more practically.

How we're seeing this manifest is the lack of continuous improvements in performance

in our traditional compute systems.

So this really spurred, you know, in 2008, 2010 timeframe, the computing community to

start saying, okay, we need to be looking to other architectures.

I think this also really led to the use of GPUs as a very common accelerator, but I think

it's also kind of spurred the interest in other architectures as well.

Now, at the same time this was happening, there's this tremendous rise in this excess

of AI and machine learning.

So as the community was thinking about, okay, well, we need to be thinking about building

new types of computers anyway to improve the performance and to continue to see feasibility

of implementing computing systems in terms of energy.

If we're going to be building new types of computers anyway, the types of workloads that

we are targeting are actually changing because of the rise of the success of AI and machine

learning.

So it would make sense to target building systems that are targeted towards these types

use cases.

The third thing that happened is the rise of the internet of things.

We're pushing compute everywhere.

So it's no longer just that compute is on your desktop or in your data center or in

your supercomputing center, but it is carried with you.

I always say on my person at any given time, I have multiple compute systems.

I have a smart ring, I have a smart watch, I have a smart phone.

So we're pushing compute out to the edge.

And with that comes pretty distinct constraints in terms of the types of performance you can

achieve, the types of energy consumption you can achieve.

And so we need to customize our compute for these applications as well.

So these three things together have sort of driven the community towards exploring brain

inspired architectures in a couple of different ways.

So there are two main types of these more kind of brain inspired architectures.

I'm going to very briefly talk about neural hardware systems before moving on to neuromorphic