Future of Federated AI and ML: Interview with Chris Piche, Founder and CEO of Smarter AI
Amanda Razani: Hello, I’m Amanda Razani with Digital CxO and I’m excited to be here today with Chris Piche. He is the founder and CEO of Smarter AI. How are you doing today?
Chris Piche: I’m doing great. Thanks for having me, Amanda.
Amanda Razani: Glad to have you on the show. So can you tell me a little about Smarter AI and what the company does?
Chris Piche: Sure. Smarter AI is a software platform for AI cameras. In the same way that Android and iOS enable smartphones, Smarter AI enables AI camera users to deploy different machine learning models (or computer vision models) to see, listen, and understand.
Amanda Razani: Can you tell our audience about your background?
Chris Piche: I’m originally from Canada, I have a background in computer science, and I’ve spent my career developing scalable video and computer vision technologies and products. including the BlackBerry smartphone and, I’m not sure if you remember a few years ago, there were fantastic commercials for a new mobile TV service from AT&T called AT&T TV. If you’re a football fan or if anybody listening is a football fan, you may remember the Manning brothers teamed up for some funny commercials about “Football on your phone.” Anyway, I was one of the people who put football on your phone!
Amanda Razani: I do remember that. So I wanted to delve into the topic of federated AI and machine learning. Can you share a little about where we are with that technology, and how that can help in digital transformation efforts?
Chris Piche: Sure. We’re very near the beginning of Federated learning. I think most people will be familiar with how AI models or machine learning models are trained today, which is by assembling a set of labeled data on a server or in the cloud, using this dataset to train a model, and then, the model can be deployed, sometimes on a server or in the cloud, or sometimes on an edge device. The limitation of this approach is that in some cases, we’re dealing or want to deal with personal or private data. There may be a challenge because – whether it’s because of regulations, best practices, or user preferences – companies can or don’t want to collect all of that data on a server or in the cloud. And that’s where federated learning comes in. Federated learning is essentially taking the same process of training an AI model or ML model, but instead of collecting all the training data in a server or the cloud, with federated learning, the model is trained at the data’s source, for example on the edge. So federated learning gives us the same result, which is an AI model that’s been trained with the same data, but through a different process where private data remains private, and it is not collected and stored by a company on a server or in the cloud. Federated learning is becoming more important as the need for privacy and the regulations around private data continue to evolve.
Amanda Razani: So this being a relatively newer technology, how would companies try to take advantage of this? How would they go about implementing this into their strategy?
Chris Piche: Well, today, we’re largely in the research phase. Academics, the big 3 public cloud companies like Amazon, Microsoft, Google, and startups like Smarter AI are all in various stages of research and experimental deployments
Amanda Razani: You had earlier spoken about some other projects you were in. Tell me a little bit about those projects. And can you share a little bit more detail?
Chris Piche: The first project I told you about was the BlackBerry smartphone. And previously, when I was still living and working in Canada, I founded a company called “Eyeball Networks,” which was a world leader in NAT traversal technology. And it turned out that NAT traversal was a foundational smartphone technology. In the early 2010s, thanks to certain enabling technologies, we all knew that smartphones would happen, with the emergence of 3G networks and touchscreen computing. t was just a question of who would figure out the product market fit first. It was Apple that ultimately figured out the product market fit and then, Android and Google followed shortly after that. While this was happening, the people at BlackBerry noticed that, as BlackBerries came to their end of life, instead of replacing old BlackBerries with new ones, customers were starting to replace them with iPhones. And at that point, BlackBerry made the decision to develop its own smartphone based on the NAT traversal technology that I had developed at Eyeball, which became one of the components of the BlackBerry smartphone operating system. During the time we developed the first BlackBerry smartphone, the stock price dropped from about $120 to $6. So I learned a firsthand lesson in how quickly a market could be disrupted based on a technology transformation.
Amanda Razani: That’s amazing.
Chris Piche: A few years ago, I saw that the same thing that had happened in the smartphone market – with Apple and Android/Google replacing BlackBerry – would soon happen in the camera market. Legacy cameras are recording video or displaying it on a screen, but now thanks to certain enabling technologies, namely neural network accelerators and machine learning models, AI cameras or smart cameras that see, listen, and understand would soon be feasible. And it was just a question of who would figure out that product market fit.
Amanda Razani: It’s amazing how a journey can lead you from that story with BlackBerries, which I remember when that was the hottest thing, we’re excited to have that. And how that journey develops into where you are now with federated AI and machine learning. And the fact that technology is evolving so quickly, really, it seems like such a fast timeline with technology.
Chris Piche: I’ve been in the software technology business for about 20 years. And looking back at when I started, a product development cycle was measured in years. We would schedule 18 or 24 months to develop a new product or product version. And, being software people, more often than not we ran over schedule. However, today we’re looking at single-digit month schedules to bring new products or new product versions to market. So everything has accelerated in the time that I’ve been involved in the industry, including the amount of time we take for each step of innovation.
Amanda Razani: And I want to ask you, too, about it’s in the new stages of research and development, but I think that security is going to be a concern too. So how do you tackle security when it comes to this?
Chris Piche: Federated learning is one of the solutions to data privacy and security. Whether we’re talking about the incumbent 3 big cloud companies or new emerging companies like Smarter AI, when companies collect our personal data, this presents privacy and security challenges, both in that data will be stored and may be shared or sold on to third parties. We’ve seen examples of this a few years ago with Facebook and Cambridge Analytica and more recently the Twitter Firehose has been in the news with Elon Musk and the Twitter bot litigation. These are examples of how our private data is aggregated, shared, and sold to third parties. And as you pointed out, there can be unintended breaches of our data privacy, such as data being hacked and shared with third parties. The promise of federated learning is that it can deliver the benefits of AI and machine learning without the intended or unintended breaches of our data privacy. Let me give you a simple example of a face recognition model. There are all kinds of practical applications of face recognition. For example, getting into a car or starting a car.; I can’t tell you the number of times my assistant or my wife has found my keys for me or had a new set of keys cut for me. Now, of course, I don’t want pictures of my face stored in a server or cloud to be sold, shared with advertisers, or end up in the wrong hands. So the promise of federated learning is that companies will be able to train the same facial recognition models at the edge, without collecting and storing my private data where it can be shared or stolen.
Amanda Razani: Absolutely. Well, I want to thank you so much for coming on and sharing about this new cutting-edge technology. And I’m sure just as a lot of other technology is coming into play too. There is a lot to learn from, and companies need to pay attention because it’ll be a massive part of their digital transformation efforts soon, I’m sure. Thank you so much for sharing your insights today.
Chris Piche: Thanks, Amanda.Amanda Razani: Thank you.