Cerebellum-Inspired Neuromorphic Computing

So I want to start with a couple concepts as we get to your work, and I want to start with asking you, what is neuromorphic computing?

Yeah, so neuromorphic means brain-like.

So what we're doing in neuromorphic computing is using the brain as inspiration for a new type of computing.

The motivation for looking to the brain for inspiration in computing is that we currently live in an era of artificial intelligence, which requires training with large amounts of data, which is requiring the construction of large data centers that are requiring significant amounts of power.

approaching the gigawatt scale.

This is something which is a big tax on the grid.

It's something which I think will ultimately limit scaling of computing, is power consumption.

Whereas the brain is orders of magnitude more energy efficient.

in the big data limit, and as a result, if we can mimic the brain, we have the opportunity to do much lower power consumption computing.

Another concept in here is a mem transistor.

What is a mem transistor for people that haven't heard of that?

So one of the concepts from the brain that we're trying to mimic is the fact that unlike a digital computer, where in a digital computer we have a spatially separated memory and microprocessor, which therefore requires data to be moved back and forth between the two, and that drives power consumption.

In the brain, memory and information processing is co-located in space.

So that minimizes the amount of data motion that's required.

So the MEM transistor is trying to do just that, have a memory function and information processing function in the same device.

So the MEM and MEM transistor stands for memory.

And the transistor is like a normal transistor, which we use for conventional information processing in a microprocessor.

So the MEM transistor combines those two functions into one device.

So I want to start with a couple concepts as we get to your work, and I want to start with asking you, what is neuromorphic computing?

Yeah, so neuromorphic means brain-like.

So what we're doing in neuromorphic computing is using the brain as inspiration for a new type of computing.

The motivation for looking to the brain for inspiration in computing is that we currently live in an era of artificial intelligence, which requires training with large amounts of data, which is requiring the construction of large data centers that are requiring significant amounts of power.

approaching the gigawatt scale.

This is something which is a big tax on the grid.

It's something which I think will ultimately limit scaling of computing, is power consumption.

Whereas the brain is orders of magnitude more energy efficient.

in the big data limit, and as a result, if we can mimic the brain, we have the opportunity to do much lower power consumption computing.

Another concept in here is a mem transistor.

What is a mem transistor for people that haven't heard of that?

So one of the concepts from the brain that we're trying to mimic is the fact that unlike a digital computer, where in a digital computer we have a spatially separated memory and microprocessor, which therefore requires data to be moved back and forth between the two, and that drives power consumption.

In the brain, memory and information processing is co-located in space.

So that minimizes the amount of data motion that's required.

So the MEM transistor is trying to do just that, have a memory function and information processing function in the same device.

So the MEM and MEM transistor stands for memory.

And the transistor is like a normal transistor, which we use for conventional information processing in a microprocessor.

So the MEM transistor combines those two functions into one device.