Show Notes
Imagine a brain on a computer chip. Jack Kendall, founder of Rain Neuromorphics has figured out how to connect artificial neurons in “neuromorphic hardware,” a brain scaffolding useful for artificial intelligence. He hopes it will make A.I. cheaper and faster for all types of applications. A fan of the Crocodile Hunter, Jack grew up in the tiny town of Belleview, Florida.
TRANSCRIPT:
Intro: 0:01
Inventors and their inventions. Welcome to Radio Cade, the podcast from the Cade Museum for Creativity and Invention in Gainesville, Florida. The museum is named after James Robert Cade who invented Gatorade in 1965. My name is Richard Miles. We'll introduce you to inventors and the things that motivate them. We'll learn about their personal stories, how their inventions work, and how their ideas get from the laboratory to the marketplace.
Richard Miles: 0:01
A brain on a chip. That is what we are going to be talking about today on Radio Cade and today on the phone I have as my guest Jack Kendall. Welcome, Jack.
Jack Kendall: 0:47
Hey, how are you?
Richard Miles: 0:48
So Jack, normally how we start this is if you could just tell the listeners what the technology is that you're working on and what it does and we'll talk later about the applications and the business model. So just tell us what is it that you're working on, what does it do?
Jack Kendall: 1:02
Yeah, we're building a new type of processor for artificial intelligence that's inspired by the brain.
Richard Miles: 1:08
Okay. So let me ask a clarifying question just for listeners who sort of aren't familiar with the AI world or the very concept of artificial intelligence, what's a useful thumbnail description of artificial intelligence.
Jack Kendall: 1:20
So AI powers a lot of things these days. I'm sure you've all heard of Siri. So Siri is powered by a type of artificial intelligence called neural networks. Google's reverse image search, a lot of really powerful recommendation engines, just like Amazon and Netflix that help you find similar products also use AI.
Richard Miles: 1:38
So AI is sort of the direction the IT industry is going in terms of making everything easier to use. Is that more or less fair?
Jack Kendall: 1:45
Absolutely.
Richard Miles: 1:45
Okay. So we're going to come back later in the show and talk about what you're doing with Rain and the overall business model, but first let me get some information about you. You know, what were you like as a kid, where did you grow up, that sort of stuff. And then maybe if you want to share about your formal education before getting into your current business. So let's start with where are you from and how would you describe yourself or how would others have described you as a kid?
Jack Kendall: 2:10
So I grew up in a small town called Belleview, Florida. It's about an hour south of Gainesville. Lot of cows and horses there.
Richard Miles: 2:17
Were either of your parents in a technical field or information technology or anything like that?
Jack Kendall: 2:22
Uh, no. My Dad was a mason. He actually started his own business doing masonry, and that sort of inspired me to create my own business as well. But he had a lot of tools in his garage and I spent a lot of time as a kid building things and playing in the garage.
Richard Miles: 2:36
And you said that one of your role models was Steve Irwin, the crocodile hunter, is that correct?
Jack Kendall: 2:43
I was a huge fan of Steve Irwin.
Richard Miles: 2:45
So between masonry and hunting for crocodiles, sort of your... that was your zone, right?
Jack Kendall: 2:49
Yeah, that's a pretty good description.
Richard Miles: 2:50
How old were you when you first got interested in the brain? When did that sort of jump out as a subject matter or concept that you wanted to know more about?
Jack Kendall: 2:58
Um, it's kind of funny. Most of my life I was interested in things like physics and chemistry, rigorous kind of deep science fields. It wasn't until I was pretty far into college that I really got interested in the brain and sort of wanted to apply the way of thinking of engineering and physics to understanding the brain. So probably about when I was around 21, 22 years old.
Richard Miles: 3:21
So what did you start out studying in college? We're you a Physics major. An Engineering major?
Jack Kendall: 3:26
Yeah, I majored in Chemical Engineering and Physics.
Richard Miles: 3:30
Okay. And so was it sort of a breakthrough moment in a particular class or a particular instructor in which you started getting curious about the brain or did this come from some other influence?
Jack Kendall: 3:40
So a good friend of mine was studying neuroscience and I had always been passively interested in the brain, but it wasn't until I read a book by somebody named Jeff Hawkins, actually the inventor of the palm pilot, called On Intelligence that I really got really, really interested in the brain.
Richard Miles: 3:58
In what timeframe are we talking about Jack? Roughly?
Jack Kendall: 4:01
Yeah, this was now about six years ago.
Richard Miles: 4:05
Six years ago, okay. So neuroscience has one of those fields that has really changed or just continues to change very, very rapidly. Is it a little bit disorienting being in a field that seems to be moving at such a fast clip in terms of what we know about the brain now is sort of multiples what we knew even probably six or seven years ago.
Jack Kendall: 4:24
Yeah. When I first started learning about the brain, there was so much information, you didn't even really know where to start. As I've studied it more and more over the years, things are beginning to make more sense and you can see that there's this broad framework that's emerging to understand the brain. So things are making a lot more sense now and I think that we're actually pretty close to a cohesive understanding of the brain.
Richard Miles: 4:48
So from the outside people who are not specialists in artificial intelligence, you seem to get one of two reactions. Either this techno enthusiasm for this is going to make everything easier and the world's going to be great because of AI and then on the other hand you get this technophobia or tech dystopia where AI going to take over everyone's job and it's going to dehumanize and depersonalize. So knowing what you do about the actual development of AI, are we closer to one of those two poles than the other one or is it like most things are just kind of right in the middle?
Jack Kendall: 5:21
This is a point at which really we have a choice collectively about how we use these technologies. There are many, many, many applications of AI that have the potential to make human life much better, especially in healthcare and preventing diseases and creating new cures for things like cancer. But at the same time facial recognition, especially in mass surveillance, has the potential to create a somewhat dystopian type reality that we may live in. So I think it really depends on what we do right now to prepare for the advent of true real AI.
Richard Miles: 5:58
So you started a company Rain Neuromorphics. First thing, tell me about the name itself, Rain Neuromorphics. What does that signify?
Jack Kendall: 6:05
So Rain, it's a play on sort of our core technology, which we're very brain inspired and we try and overcome some of the scaling problems with other AI chips by making these random networks. So Rain stands for Random Artificial Intelligence Networks and then Neuromorphics is a combination of "neuro," which means "brain," and "morphic," which means "in the shape of." So we're building things that are in the shape of the brain.
Richard Miles: 6:31
So Jack, for lay people, how would you explain, I guess the competitive advantage of the type of work you're doing versus other folks working on AI?
Jack Kendall: 6:39
Basically, we can build larger networks, larger brains than any of our competitors and train them faster. So right now a big problem with neural networks is that it's very difficult to train them because the amount of time that it takes to train scales as you make the network larger and larger. Right now that scaling is very poor, but we have an architecture that scales more like the brain does, so it's much better in terms of as you make the network larger and more powerful, it still is easy to train.
Richard Miles: 7:12
And is there something distinctive about the type of network, that is explainable, that gives you this advantage.
Jack Kendall: 7:19
In conventional architectures, what you see a lot is you see... basically in AI right now it's all about neural networks and in particular something called deep neural networks and these are computational algorithms that work in a way that we think is really similar to what the brain is doing and they're powerful as a function of the number of neurons that you can really fit into the network. And so conventional algorithms and hardware, you grade these really dense, they call them fully connected, networks. So every neuron in a layer is connected to every other neuron in another layer. But the brain isn't like this. The brain doesn't have these fully connected grid-like networks. In the brain, things are very sparsely connected. So a neuron might only connect to maybe less than one percent of its neighbors. So this is called sparse connectivity. It's really hard to implement in hardware, but we have found a way to implement this sparsity in hardware and that is what allows us to scale.
Richard Miles: 8:18
What are the applications of Rain Neurmorphics and who are your clients and what are you working on for applications?
Jack Kendall: 8:25
Yeah, so what we want to do is replace a lot of the compute devices that companies like Google and Amazon are using right now for their AI. Most people right now are using graphics processing units or GPUs, because GPUs were the best solution that already existed when neural networks became really popular. So what we're trying to do is replace a lot of those devices with special purpose hardware for neural networks, for very large companies that do a lot of AI.
Richard Miles: 8:53
Are you licensing this for someone else or do you own the patent to this?
Jack Kendall: 8:56
Yes, so the University of Florida, it was developed at UF. So UF owns the patent and we have a full exclusive license from UF.
Richard Miles: 9:04
And is the idea to do a sub-license to what these other companies or... to get acquired by them or what is your thinking in terms of...
Jack Kendall: 9:11
We want to build chips. So we want to really build the chips and sell them to the end user.
Richard Miles: 9:17
Okay. Um, how big is your company now, Jack?
Jack Kendall: 9:20
We're pretty small. We're still an early stage company. We just recently raised our seed round. We're hiring pretty fast right now, but right now we're at about five people.
Richard Miles: 9:28
So as you set out to explain this technology to your clients or potential clients, was it a bit frustrating? Did you have people not quite get it or maybe they understood how the technology worked but they didn't really see the value proposition in the application? Or has this been one and done you make your pitch and they sign up?
Jack Kendall: 9:44
At first people were skeptical. Um, we're doing something that's very different from traditional silicon design, traditional processor design. So they should have been skeptical. But pretty rapidly we built up a very good list of people who really know the technology very deeply and can vouch for it. And once we had some big name people that were on board, convincing others became much, much easier.
Richard Miles: 10:10
So you talked about one sort of very useful "bad experience" you had in the sense that apparently you're applying for a grant and one of the reviewers said they liked the idea but you didn't have any useful learning algorithms then and it was essentially like a brain that didn't know how to learn and you said you took that rejection pretty hard, but it sounded like out of that rejection came something good. Can you talk a little bit about that?
Jack Kendall: 10:32
Yeah, definitely. So I remember this very well. I was doing research at UF under a professor Dr. Juan, you know, at the time and he's a material science professor and this is where I came up with the idea and we submitted the patent and we needed funding. So we were applying to all these grants and it was multiple rejections in a row and that is kind of demoralizing and I really did take the feedback from the reviewers close to heart. And one of the things that was obvious was that this was really just a framework architecture. We didn't have any algorithms, so it was like we had built this hardware, but there was no software for it to run. So after I got that feedback, I was like "All right, this is not gonna go anywhere unless I can build really the software for it." And so I became kind of obsessed with this problem and really just thought hard and long enough that I came up with something that worked and after that that's when things really started to take off.
Richard Miles: 11:27
So this brings up an interesting question because for a lot of inventors and entrepreneurs, there's a certain amount of criticism that you kind of have to ignore, right? Because if you took into account every piece of criticism, you just stop, but at the same time there's criticism you have to pay attention to because otherwise you're never really going to progress. Was there something about this way, this advice or criticism was delivered that you realize that this is the type of stuff I need to listen to and not ignore?
Jack Kendall: 11:51
Yeah. This has happened a few times, but it's always the criticism that makes you deeply uncomfortable. Because you know that it's true.
Richard Miles: 12:01
I see.
Jack Kendall: 12:01
The things that you know are irrelevant and doesn't do anything to you emotionally, that's usually the type of criticism that you can ignore.
Richard Miles: 12:09
Yeah.
Jack Kendall: 12:09
But the ones that really make you think and question what you're doing, that's the criticism you have to pay attention to.
Richard Miles: 12:16
Right, because they've gotten at a root problem that you... you acknowledged, you just didn't want to fully acknowledge, I guess.
Jack Kendall: 12:20
Exactly, yeah.
Richard Miles: 12:22
Interesting. I had another guest recently. I asked for advice and he said, always tell yourself the truth because if you don't, you'll sail a long boat, believing all the very good press as they say. So what are your immediate next steps, Jack, aside from cutting a huge deal with Google or Apple or something, are there immediate hurdles that you're trying to overcome? Either regulatory hurdles or financial hurdles or whatnot.
Jack Kendall: 12:45
There's not much regulation that we've run into just yet. Financially we're doing okay right now. So our main challenge is actually putting out our alpha chip. So we're scheduled to have that out in about 9 to 12 months. So, we're starting our design process. We're hiring all the engineers that we need to do that, but it's going to be a long journey, but we think it'll be worth it.
Richard Miles: 13:06
And does rain have a website that people can go to if they're looking for more information or...
Jack Kendall: 13:10
We're actually in the process of setting that up right now. We have a landing page. It's rain-neuromorphics.com. We'll have some more stuff up there very soon.
Richard Miles: 13:19
Well, great. Jack. Thank you very much for joining us today and we look forward hopefully to getting updates and then having you back on the show.
Outro: 13:28
Radio Cade would like to thank the following people for their help and support. Liz Gist of the Cade Museum for coordinating and inventor interviews. Bob McPeak of Heartwood Soundstage in downtown Gainesville, Florida for recording, editing and production of the podcasts and music theme. Tracy Collins for the composition and performance of the Radio Cade theme song featuring violinist Jacob Lawson. And special thanks to the Cade Museum for Creativity and Invention located in Gainesville, Florida.
