Manufacturing without process management….coding without software engineering…construction without blueprints…throughout human history we’ve realized that there’s a difference between core capabilities and the “art of the applied”. We’ve learned that, until we develop a methodology of practical application, our success remains limited.
Artificial Intelligence (AI)—including Machine Learning (ML)—stands at this same precipice today as it transitions into the mass market. Yet because we are only beginning widespread commercialization of AI, there is a missing body of knowledge—you might say an unwritten book on the shelf. This is a body of best practices that are essential to developing AI applications with low risk, high value, and high adoption, across the thousands of use cases in which the technology can provide massive value to humanity worldwide.
There has been an active community of researchers supporting applying AI in specific cases, as showcased in the long-standing series of conferences on Innovative Applications of Artificial Intelligence (IAAI) and forums covering applications in more specific domains, such as Machine Learning for Healthcare. However, such discussions tend to focus on each individual application project in isolation both from other applications and from the total business context in which the application is successful, or not.
I and a few colleagues have spent the last few decades exploring this space from a more structural technology-transition perspective: transitioning AI from its academic and research roots into practical applications. Along the way, we’ve encountered obstacles and challenges that were unforeseen by the scientists that invented much of what we know as AI today. And, as we step back to analyze successes and failures, we’re starting to observe systematic patterns: traps for young players, subtle but pervasive challenges at each stage of a project lifecycle, as well as expert best practices. Sharing them with you is the purpose of this blog.
A bit about me: I’m a computer scientist, entrepreneur, investor, and advisor who’s driven innovation in almost every aspect of AI—from foundational science to applied enterprise systems—for more than 30 years. Because in my work I’ve been accountable for successful end-to-end applications, I’ve had the chance to think deeply about what separates failure from success. I’ve also been honored to enjoy a front-row seat to over 100 AI projects: watching the “magic” of AI leading to innovation breakthroughs in many sectors, including semantic search engines like Powerset and Bing; agriculture; government work across all enterprises of NASA; multiple investments in AI-enabled healthcare, pharma, and medical technology; transportation, retail, consumer and entertainment, lunar payloads, and more. Also, as a founding board member at Singularity University, I have worked to advance thinking about AI technology and applications for organizations beyond my own around the world.
There’s a lot of hype around how AI systems are going have a big impact, and analysts are projecting growth rates of AI worldwide well into the double digits year-over-year. Yet a growing body of evidence suggests that, like any discipline applied to a wide range of new problems, many (some say nine out of ten) AI projects will fail. Although these growing pains are typical of any discipline “crossing the chasm” into the mass market, we must still face them head-on. Those who succeed in this market won’t necessarily be the scientists with the fastest or highest-accuracy algorithms, but the ones who design and build systems that actually work for all stakeholders in the full context in which they will be marketed, evaluated, selected, implemented, adapted, tested, deployed, integrated, adopted, used, managed, maintained, and trusted. And AI deployments usually aren’t either a total success or a total failure. There’s a wide spectrum of outcomes, and evolution over time, so it’s important to understand what’s good enough, and to build on that knowledge, to keep learning and improving.
The secrets to this kind of success are not written down yet: there is a huge knowledge gap between what’s possible and what’s successful in the application of AI. Indeed, applying AI successfully is often harder than the core technology, and is much less understood. This blog is about filling this gap, and increasing your chances of success.
And what’s more, with a little luck, hard work and plenty of creative insight, we’ll start a dialogue and build a community of members who will collectively move the “art of the applied” forward to help organizations of all kinds solve problems, bring unique solutions to market, and meet the needs of all critical stakeholders, while maximizing the societal benefits and minimizing the harms of this powerful new technology.
As we go forward, I’ll be keeping my finger on the pulse of new AI developments; sharing my experiences and expertise; and providing guidance on how to bring AI applications projects to fruition. Along the way, I will explain a framework for practical approaches to the application of AI for project success and ultimately, the entire innovation journey, from market discovery to successful exit.
Please consider subscribing to the blog for regular insights and updates. My commitment to you: a regular stream of fresh and practical ideas you probably haven’t heard anywhere else, and which will lead to your best shot at a breakthrough AI success of your own.