The Business Leadership Podcast

Salli: [00:00:00] You're listening to the business leadership podcast with Edwin Frondoso.

Bob: I say brand new, six years old or so, but it's totally transformed the amount of compute required to get the job done and using traditional techniques from the CPUs from Intel, GPUs from NVIDIA, You're going to end up,

you know,

burning down the planet if I

could be hyperbolic.

Bob: AI is going to be deployed on all of those platforms and everything in between and running them, which is what inferences that's going to be done on those machines in those locations. And so what we're trying to do is really change the equation about how much energy and how much cost it is to run these neural networks

Edwin: Good morning. Good afternoon. And good evening biz leader. Welcome to another episode of the business leadership podcast. I'm your host, Edwin Frondozo. And today we are [00:01:00] featuring. A special episode. From our future narrator mini series recorded live at the collision conference in Toronto, Canada.

In this series, we explore the future of leadership, innovation, and storytelling. With visionary leaders who are not just designing products. But our creating entire new worlds and markets.

Joining me is Dr. Paul Newton and together. We are speaking with Bob Beachler. He is the chief product officer at Untether AI, a Silicon valley veteran with extensive experience in the development and marketing of FPG AEs software tools. Vision processors and AI acceleration devices.

In our conversation, Bob will discuss untether AI's mission to solve energy crisis in artificial intelligence, computing. We'll explore how traditional computing methods like CPU's from Intel or GPU's from Nvidia are energy intensive and unsustainable for AI workloads. Bob will explain how untether AI aims to make [00:02:00] AI energy efficient. Significantly reducing the power required for neural networks by nearly an order of magnitude. This transformation has broad implications from data centers to everyday technology, such as cell phones and autonomous vehicles. So without further ado. Here we go

We're now speaking with Bob Bleacher, the chief product officer at Untether AI.

Bob, how are you doing today?

Bob: I'm doing great. Thanks for taking the time to talk to me.

Edwin: Thank you for dropping in. Just jumping right in, I'd love it if you could share what problem is Untether AI solving today?

Bob: Yeah, so what we're solving is the energy crisis in artificial intelligence. AI is a brand new compute workload.

I say brand new, six years old or so, but it's totally transformed the amount of compute required to get the job done and using traditional techniques from the CPUs from Intel, GPUs from NVIDIA, You're going to end [00:03:00] up, burning down the planet if I could be hyperbolic. And so this company was founded here in Toronto Really to focus on how do you run these AI workloads in a more energy efficient manner?

Which goes to your total cost of ownership goes to sustainability Ecology and the so that's what we're really focused on

Edwin: Just thinking about my tech background when it comes to you know Even building a four 86 computer or three six. when you talk about energy consumption, this is for the benefits of the business leaders who may not.

So like how significant changes are we looking at that on tethers doing in terms of this, the savings?

Bob: Oh, it's immense. So we're reducing the amount of power necessary to run a neural network by almost an order of magnitude. And we're talking about. Gigawatts of power. Once AI is deployed in the world at scale, we're going to see AI touch almost every aspect of the technology stack [00:04:00] from the data center all the way to your cell phone all the way to satellites.

AI is going to be deployed on all of those platforms and everything in between and running them, which is what inferences that's going to be done on those machines in those locations. And so what we're trying to do is really change the equation about how much energy and how much cost it is to run these neural networks.

Edwin: That's amazing. That's super exciting. And AI, artificial intelligence, it's really the up and coming, the new thing. I'd love it just from your point of viewYou mentioned AI is going to touch everyone, but how do you see it really revolutionize the everyday world?

Bob: Yeah, so I think two things. One is it's solving problems that couldn't be solved before with traditional programming methods. So deterministic programming can only get you so far. The thing about neural networks is they teach themselves to solve these problems. So things like [00:05:00] autonomous vehicles, you couldn't do all the object recognition using regular programming.

You have to train neural networks to do it. Same thing with large language models and generative AI. You can't just program the entire internet into a traditional C program. You create these giant large language models and then train it on the entire known sum of knowledge in the internet. And now it has that codified in there and you can ask it questions and get responses.

So it's solving those types of intractable problems that you can't do with normal programming. And then how that translates into the experience that you and I have is that it's going to be a productivity enhancer. You're going to be able to do summaries, for example, of this podcast. If somebody just wants the salient points, shove it into a large language model and say, Hey, Tell me the salient points, and it'll give you an outline of what was discussed.

So it's really about making humans more productive in [00:06:00] conjunction with AI. No, amazing.

I'm thinking about, and it's pretty wild. It's just changing so fast. And we're talking about saving energy. We're talking about enabling everyone. I'm wondering what are you seeing as well within the industry?

What are, what other disruptions are happening within the industry?

Right now we have a singular disruption going on, which is the training of these large language models. They're so large, and they require so much compute, and it's in the headlines every day. NVIDIA is the world's most valued company as of yesterday, and that's on the backs of training.

They're selling hundreds of thousands of these GPUs to do the training, and there is a short term disruption in the supply chain. And so people are spending money just to get on allocation to get those GPUs to do the training. That's the disruption that's happening right now, but what's really going to disrupt is when you're done with your training, now I want to deploy these models [00:07:00] and I'm going to have millions of people asking questions and giving response, that's the inference portion and that's the next level of disruption of what, we're aiming for in terms of providing that efficient compute for that deployment.

Edwin: And when we're talking about deployment in that second phase, that's, we're talking about having access to these large Langle models for everybody to have access to. And we can't even, maybe you have a lot of insight on this, like what that really means and what

Bob: availability even, right?

Availability, one of the things that we're seeing and the Canadian government has done some stuff about this, about sovereign compute, but we're seeing this in every country now. Where they're doing things specific to that country to optimize, for example, large language models for that area, training it in their own languages.

So a company here, Cohere has been working on generative AI, and they're, have a big focus on making sure that they're handling all of the languages in the world. I'm not just [00:08:00] training on English or Chinese. They're doing small, languages that, don't exist there. commonly places in Africa and the like.

So you're going to see all of that. So it's not going to be one model. There's going to be millions of models all tailored for specific applications in specific regions.

Edwin: That's wild. It just blows my mind. And I was thinking to myself, Oh, there's thousands of models. How many models do I get access to?

And really help us get better.

Bob: And the thing is that as a user, you probably won't even know it. It's going to be under the guise of your UI on your iPhone or on your, your Android phone or in your car or on your computer in your zoom, it's all going to be there just as another tool, but it's all running AI in the background.

It's just going to make you more productive. That's great. Bob,

Edwin: what does a future of quote faster and more cost effective AI look like? From the cloud to the edge, end quote, look [00:09:00] like.

Bob: Yeah, and so that's one of our taglines. And that's because our compute architecture is scalable. We make chips that are big enough to run in data centers, but also small enough so that we can put them in satellites in low earth orbit.

Or put them in cars. Or put them in tractors. So that, because, these models, they'll, and like I said, there's thousands of them, they'll be deployed at different points. across the entire stack. And so for us, we take our energy efficiency and apply it in data center applications, in autonomous vehicle applications, in aerospace applications, smart city, smart retail.

The one thing we don't do is we're not in your cell phone. How come? Because that's so cost constrained, and they're going to make their own integrated CPUs with a small amount of AI processing. To use a good example, engineering jargon, there's something called tops Tara operations per second.

Your typical cell phone processor will have about five, zero [00:10:00] tops of AI compute performance. Our single chip has two petaflops, so 2000 Tara flops, right? So we're talking about, deployment at scale. We're not doing a single camera on a cell phone. We're going to be aggregating, the 30 sensors and cameras in your car or on a tractor or.

In a smart city application where there's thousands of cameras and you want real time, actionable intelligence. It's amazing.

Edwin: When it comes to leadership, when you're talking about product working with, your employees, stakeholders, clients, I'm curious, what's your biggest challenge right now?

Bob: So for us, the challenge, I think in the industries we talk about is the explosion of models. And there's just so many of them and there's new papers out every day. And so our challenge is to make sure that what we create both on the silicon and the software side can adapt and support all of the new innovations that are happening.

So from [00:11:00] a business and technical standpoint, that's one of our biggest challenges. From a leadership standpoint and leading a company, Because of the space we're in, it's so exciting. It's so vibrant. There's so much change. People are excited to get up every day and go to work. I am.

And I've been making computer chips for 35 years, right? And it's just I've never seen anything like this since back in the early days of the semiconductor industry and the startup world. So the energies there, the excitements there, it's managing some of the chaos and keeping people focused on, moving forward and completing the vision.

Edwin: Yeah, so how do you plan and alleviate for those excitements and the chaos?

Bob: Yeah, so two things. Making a computer chip takes a long time. We're typically two years between generations. We've done our first generation chip that's already deployed. Our second generation is now in the lab.

And we're only six years old. So that planning starts now. So we have our second generation in the [00:12:00] lab. We're finishing up the definition of our third generation right now. And the design is ongoing, right? But that's one cadence, the software cadence, which is how do we get the model to run efficiently on our Silicon?

That cadence is a three month period. So that's agile development scrums. We have planning before each three month sprint. For the next release, and that gives us the flexibility to adapt to say, okay, now we're seeing this shift in the industry, we should work more on natural language processing, because we've done all of the vision networks that we need to do, and so that gives us flexibility there.

Edwin: Yeah. Five minutes, thanks. Bob, what is, This is more on the personal side, but what is your vision of the future that you are building within, on Tether AI?

Bob: Yeah, and it goes back to what I said previously, which is our vision is that we become [00:13:00] the inference solution. When you're running and deploying your models at scale, we'll be able to run those models faster, cooler, and cheaper.

Whether you're in the data center, whether you're on the factory floor, whether you're in an autonomous vehicle, in a vision guided robot or, like I said, even in satellites. Because AI will be deployed in all of those different platforms. And so our vision is that we become that solution of choice for running these models.

I

Edwin: guess I'm curious, follow up to that, When Untether AI becomes the solution of choice, what does the world

Bob: look like? Think the world looks like it's going to be an AI powered world. AI is going to make people more productive. Your, agricultural will be more productive.

You won't be wasting as much pesticides. and herbicides because you can do more targeted types of farming. [00:14:00] You're so you're going to be more productive in your food processing. You're going to be requiring less power so that you're more green and ecological in terms of sustainability. And you're going to be more productive because AI will be used in every aspect of your life to make you more productive.

Paul: And so what changes for humans and just life in general? Like, how is it going to look

Bob: for us? Like I said, it's going to be integrated into your daily life, whether it's making travel plans, whether it's going to the grocery store, AI is going to assist you in every step of the way. I'm a believer personally, right?

I'm not an AI doomer. I believe it's going to make us better. It's not going to replace jobs. It's just going to make you more productive. And therefore it's going to be a huge economic benefit. This is a seminal change. Like the invention of the microprocessor. Like the invention of the [00:15:00] Internet.

And think about what the world was like before the Internet and the World Wide Web. A lot of people don't. I'm old enough that I remember. Me too. And it's just everything is online now. And then, in the future, everything's going to be AI system. And you're just going to be better. And

Paul: personally, like how has life gotten better for you?

Like what, like what experiences do you have now that weren't possible?

Bob: Oh, geez. Just even the thought of working with companies on autonomous vehicles, 20 years ago that, that was a pipe dream. Now it's absolutely a reality. Text generation. So my marketing team is more effective.

Because they can, use tech generation to create marketing collateral. Your research is much more effective because you can scan all of the papers and understand what's going on. It's all to me about that productivity.

Paul: And do you see it having like more demand on people or less? There's just so much more coming at us.

Like being [00:16:00] more productive, does that mean We're way busier and have to look at things coming from all angles or

Bob: what? That's the thing is that if the AI gets good enough, you should be able to deploy it so that you don't have to pay attention to it. But you have to get to that level of trust, which we're not there yet.

I'll get, I'll pick on autonomous vehicles. They've had a series of setbacks with crews and some other folks. It's it's not quite there yet where I can just deploy it and forget about it. I have some advanced driver assistance in my car. I don't trust it yet, and I'm involved in the industry.

I know who makes it, right? And it's just yeah, you still gotta keep your eyes on the road, guys, folks. It sounds like

Paul: the clincher there, that trust, or like that bridge where we have to just be like, just give up and embrace. And

Bob: one of the great things is that the human mind, we're not even close with artificial intelligence to what the human mind can do, and the adaptability of the human mind.

I think that when you see this generation coming up that's living with AI, right? And what they're, how they're going to function 30 years from now is [00:17:00] totally different than how we're functioning today. Because they're going to live in an AI powered world.

Paul: Do you have any inkling of what that might even look like?

Bob: I don't, and that's the beauty of, I think the human mind is that our adaptability to change. We talk about business leadership and what's the chaos of being in a startup in such a tumultuous industry as AI. We're adaptable and we can change AI and to get to AGI, artificial general intelligence, building in that adaptability is super hard and it's not something that, that the AI community has even really, some people are starting to look at that, but it's a long way from fruition.

Paul: But it's like paradigms are blowing up everywhere now. 'cause we just can't even think along the same lines, the way things have been done.

Bob: Yeah. Yes. It's a matter of being adaptive. Yeah. And I think that, some people will be able to adapt, some people may not.

I'll just use my family generationally as as an example, my father's never used a cell phone. [00:18:00] How does that happen? How is there someone still alive now? He's 95 years old. Yeah, but he's never used a cell phone. He's never used a computer. Myself, I started using a computer at 14. I wrote in, I used Unix and Linux.

And it's just to me, technology is second nature. And I've adapted to different user interfaces and all of that type of stuff. But that's how I was brought up. And I think this generation coming up will be that adaptable.

Paul: Oh, that's so exciting. I'm just going to hand it back to Edwin. I want to talk forever, but

Edwin: I think that's it.

I think we're time, Bob. It's been amazing, but I just wanted to thank you for joining us on the business leadership podcast.

Bob: No, I really enjoyed it. Thanks for having me.

Edwin: Amazing.

That's it izy leaders. Thanks for joining me on this special episode of the business leadership podcast. Part of our future narrator miniseries recorded at the collision conference. This was an amazing conversation with Bob Beachler exploring how untether AI is revolutionizing energy [00:19:00] efficiency. In AI computing.

For links to all the resources that we discussed to connect with Bob, and to learn more about the future narrative projects, please do swipe into the show notes in the app that you're listening to right now. And if you are interested in reading more about Bob. And the other amazing business leaders that we profiled, please do join the waitlist for our upcoming book.

And by the way, if you found value in this episode, please subscribe, rate and share it with the very first person who comes to mind that could benefit. And be grateful from hearing from you.

Your support helps us grow and bring you more great content. Thanks again for tuning in and being a part of our community until next time have a 100 X day.