Jags Kandasamy, Latent AI: “most enterprises simply lack the expertise to successfully implement AI”
While Artificial Intelligence and Machine Learning have been around for some time now, enterprises still struggle to harness the power of these technologies.
As computing power shifts from the cloud to the edge, AI projects are under even more pressure to perform and derive meaning from the incredible amounts of data now available to them. While the technology is advancing at rapid pace, yet, many AI projects still fail.
To delve a little deeper into the challenges that companies are facing in terms of adopting AI and ML technologies, Cybernews reached out to Jags Kandasamy, the CEO and Co-founder of Latent AI, an edge AI implementation and workflow services provider.
Let’s go back to the very beginning of Latent AI. What has the journey been like over the years?
First, thanks for the opportunity to talk about Latent AI and the future of edge computing. Although we are a new company, Latent AI has deep roots in creating industry-defining technology. We initially spun out of SRI International, the birthplace of Siri, among other innovations. We started Latent AI because we knew the world of computing was changing and AI was going to play a major role in that transformation. What we underestimated were the challenges in getting to where we are now. A lot of that was simply because the technology and what it can do is still being defined. But it’s been a fruitful journey, and one we are glad we are on. We are thankful we have found fellow travelers in investors and team members who are as dedicated to creating a sustainable and vibrant future driven by edge AI as we are.
At Latent AI, Edge AI is the main focus. Can you briefly explain what this field entails?
Absolutely. Edge AI is the natural evolution of AI. What we are seeing is the computing power of the cloud transferred to the edge. That’s a necessary and fundamental step in how enterprises will eventually need to handle the massive influx of data new technologies are delivering to them at an ever-increasing rate. Think of how many cameras are out there now that weren’t even ten years ago. Everything is collecting data now, no matter its original design.
Edge AI solves a lot of the problems the data explosion causes. What we are doing at Latent AI is helping enterprises move the ML model to the data, instead of the data to the model in the cloud-based AI model. That lessens power requirements and is inherently more secure since the data doesn’t have to be transmitted for processing.
Can you introduce us to your Latent AI Efficient Inference, or LEIP, platform? How is AI incorporated into your product?
LEIP is our edge AI development and deployment platform. It gives ML Engineers and AI developers what they need to build edge AI ‘software factories’ that automate and simplify optimized model delivery. All our customers need to bring is their data. LEIP does the rest with automation that dramatically reduces edge model training time while also maintaining accuracy. What is unique about what Latent AI does is that we combine model compression plus compilation into an automated software factory that produces optimized models for specific architectures, parameters, and hardware. That really makes it easy for enterprises to move their model into production because LEIP produces executable runtime engines with all its dependencies pre-packaged. And it's delivered in a container along with integrated health and security checks, so it can run anytime, anywhere, and at scale.
How have recent global events affected your field of work? Did you add any new features as a result?
We are in a rapid and historic period of change, for certain, and not all of it for the better. Some of our newest features will be what enables enterprises to better adapt to that changing world. For technology, events are reinforcing the need for better security mechanisms. The war in Ukraine is certainly a cause for concern about nation-state-sponsored cyberattacks. Once that genie is out of the bottle, it won’t go back in. As we mentioned earlier, Edge AI is inherently more secure because the data stays on the device, and doesn’t have to be transmitted to the cloud for processing. We will always be looking to add more security features to what we do as a matter of course and have plans in that area.
On a much larger scale, the truth of climate change is also getting harder and harder to deny, as are rising energy costs. It’s having an impact already, and it’s only going to get worse. Edge AI is built for the modern world in a way cloud-supported AI isn’t. It’s inherently greener because it uses less power and has much lower computing requirements. And it can save enterprises money while also reducing some of their carbon emissions.
Since AI is a relatively new technology, people still tend to have some misconceptions and myths regarding it. Which ones do you notice most often?
The greatest misconception is unfortunately that ‘we are going to get it right.’ AI is not easy to implement. Most enterprises who attempt it are so lured by the promise of AI that they rush past how littered the landscape is with projects that have failed. According to Gartner, only 53% of projects make it to production. In fact, a full 85% of AI projects fail to meet their initial business goals.
So why is it so hard? Most enterprises simply lack the expertise to successfully implement AI. The industry hasn’t existed long enough to create the experts necessary to expand it yet. And even if enterprises have experts at their disposal, they often lack the computational infrastructure to support the project.
The other misconception, or maybe misunderstanding is a better way to frame it, is data. Enterprises simply don't understand the sheer amount of data they have now, let alone the tsunami that is coming. While that data presents a considerable management problem, it also presents a business opportunity as enterprises can gain tremendous insight from their data by applying machine learning to discover insights too complex for a human to recognize.
In your opinion, which industries should be especially concerned about implementing AI solutions?
The technology is still so new most industries have yet to figure out all its potential applications. Any industry that can leverage object detection can benefit, of course. Implementing quality control on assembly lines, identifying humans, and improving health care monitoring are only a few of the possible practical implementations. It’s basically whatever you can imagine at this point, especially in the field of object detection. What’s really going to separate the wheat from the chaff are the individual companies that can learn how to leverage edge AI first. They’re going to gain a competitive advantage over their competitors that will be hard to catch up to. The organizations that can establish best edge AI practices are going to be able to shift their strategies far more effectively and get an opportunity to pull ahead and stay there. And it’s those potential gains and their incredible benefits and insights that are driving the technology in the first place and making enterprises take the risk of trying.
Right now, we are working with a wide cross-section of industries such as the agriculture industry to solve weed detection issues so they can use fewer pesticides. We are working with customers to improve their factories to produce goods more efficiently with less waste. And we are working with companies who are deploying large-scale IoT devices that can help us better understand our environment. What truly excites us is imagining all the new ways our technology can be applied. We are only at the beginning of what edge AI can really do.
What predictions do you have for the future of AI technology?
Edge AI is poised to explode. There are simply too many forces pushing that reality into being, from the limitations of the cloud to climate change. What we are doing is making sure enterprises can capitalize on edge AI by giving them what they have been missing - a repeatable path to success. By automating model development and deployment, we can speed the delivery of optimized edge models while making the whole process simpler to manage. And by resolving those edge AI deployment challenges, we free enterprises to instead focus on their business outcomes.
While LEIP helps enterprises build their own edge AI software factories, we also knew we wanted to go further to close the gap between edge AI expertise and project implementation. We created LEIP Recipes to solve deployment and maintenance challenges and give enterprises a better way to implement edge AI. LEIP Recipes are a set of pre-configured assets combined with a set of instructions to follow to get to an optimized model. Each LEIP Recipe tackles a type of problem like object detection or classification and is configured for a specific model and hardware target. It comes with all the model optimization settings pre-configured including quantization, structured pruning, and throttling, and lets ML engineers leverage models pre-optimized for low power, inference speed, size, and memory. We believe Recipes are the missing piece for enterprises who want to implement edge AI in a better, faster, and more reliable manner.
In this age of ever-evolving technology, what do you think are the key security practices both businesses and individuals should adopt?
Edge AI follows the cardinal rule of cyber security – build it in, don’t brush it on. Security is simply an inherent part of edge AI because the data stays on the edge. That solves a whole lot of security problems. And that’s what organizations need to both understand and explore – that the old ways of security are barely holding together. Look no further than how all the new password complexity requirements have become self-defeating and caused people to just re-use the same one across multiple sites. When one is breached, then they all are. So enterprises should always be looking at solutions that build in security and that can be resilient against the most common methods of attack.
The password problem is going to be solved by a combination of partnerships like the one Apple, Microsoft, and Google just announced – but also by what can be powered by edge AI. Standards like FIDO2 are going to transform authentication over the next decade and hopefully let the security industry get a handle on threats like phishing. That means building better bio-authentication mechanisms to support those new standards. Edge AI will be a critical part of that. Again, part of the problem with current authentication methods is the transmission of the data. Edge AI is going to let organizations do it all on the device itself and do so in fraud-proof ways. That will be a welcome change for all of us.
Would you like to share what’s next for Latent AI?
We’re excited about LEIP Recipes, which we discussed a bit above. We are going to continue to help enterprises deliver on the potential of edge AI by providing the tools, support, and expertise necessary to make it happen.