Capturing The Value in AI is
More Elusive Than People Think
Harnessing the Power of AI Will Require
Companies to Choose Which Capabilities To Invest In
Scroll to access the full insight
Marvin Minsky, a cognitive scientist and one of the fathers of artificial intelligence, liked to say [in the early 2000s] that we can make a computer capable of beating the reigning genius of chess, but we can’t make a robot capable of walking across the street as well as any normal two-year-old child.
The real world is not a strictly regulated, closed system like a chess game. Sensing a player’s moves on a wired chessboard and responding quickly and intelligently is one thing. Sensing—and physically responding to—reality (stones, curbs, potholes, pedestrians, oncoming cars) is quite another. In fact, the entire AI industry has been through multiple attempts since the 1980s to grow into a mature market, but each time collapsed because it was unable to meet the unrealistic public and investor expectations generated by non-real-world computing triumphs like those of IBM’s Watson.
Of course, digital computing has radically transformed human affairs. But so far, that transformation has taken place on the computer’s terms. The marvels of AI and machine learning have largely taken place in rigidly regulated, closed systems; so far, it’s been about how many guys in white lab coats do I need to make AI tangibly valuable?
Rapid advancements in silicon, computing and networks are clearly forming the foundations for machine learning capabilities to advance, but these systems are still in their infancy. Many intended use cases for AI can be accomplished with less sophisticated, more cost-effective and traditional tools such as basic regressions. Value derived from machine learning and AI will be realized unevenly across markets, applications and use cases.
AI REQUIRES CAREFULLY ALIGNING NEW TOOLS to INTENDED APPLICATIONS and USE CASES
Multiple parallel technology developments have evolved that appear now to be increasingly reinforcing and accelerating one another. Cloud infrastructure resources are providing unprecedented computing scale. Mobile computing devices are extending the reach of computing. Embedded systems and IoT technology are connecting and integrating a broad array of physical and digital applications. And, the ability to capture and process massive amounts of data has the potential to inform many new and diverse capabilities.
Each of these technologies is powerful on their own, but creative combinations of these capabilities are multiplying their impacts. Human-connected devices and machine-connected IoT devices enable exponentially more data. The scale of infrastructure [cloud] computational capabilities then enables us to capture and analyze all that information. Which, in turn, sets the stage for AI and machine learning tools to analyze and capture new insights.
This new chapter in the evolution of machine learning and artificial intelligence is motivating tech developers and users to search far and wide for relevant applications to apply new advanced neural nets and deep learning tools. Most companies believe that implementing advanced AI solutions will lead to significant efficiencies, growth and competitive differentiation. However, matching new tools to high value applications and use cases will challenge many industry participants.
We all know that AI needs large data sets in order to learn, but what most people do not understand well is the actual scale of data required. While many people have come to understand that AI systems require more information than humans to understand concepts or recognize features, they don’t understand that many AI applications will require thousands or even hundreds of thousands times more data. If you examine applications where machine learning is successful it quickly becomes apparent that the bulk of effective applications are in domains where the user can acquire lots of data – think facial or speech recognition and then think about technology developers like Apple, Google or Facebook who all have vast troves of data they can access and work with.
An additional challenge beyond getting access to data is the fact that most machine learning applications today can only address a single type of learning. There really are no neural networks in use today that can be trained to address multiple parallel applications such as identifying images, playing video games, predictively diagnosing machine failures and listening to and identifying music. The working systems in use today only address “narrow purpose” applications.
We believe the domains and applications that are primed to realize significant ROI from the application of advanced AI solutions have distinct characteristics. Based on our research and consulting work, we believe there is a discernible set of attributes that characterize those segments and applications where the impacts from new AI tools will be higher and more straightforward to achieve and where the user’s propensity to adopt new tools, methods and approaches is also higher.
Domains with applications that contain diverse equipment and systems that have over many years been subject to continuous and multiple measurements, where the capital cost of the equipment tends to be high and where the economic impact of the equipment and systems “in-use” is high and where fractional improvements or cost optimization will yield significant business results. In these applications, users traditionally have understood the value of monitoring and collecting data. These markets and opportunities are typically more “mission critical” in nature.
Additionally, domains and applications where customer data is actively collected sets the stage to combine these data sets with adjacent applications to address new opportunities such as customer prediction analytics. New opportunities are emerging all over the economic landscape to fuse very large data sets with new sensor readings and measurements such as combining weather forecasts and legacy weather models with new sensors on vehicles that take real time road condition measurements to help large fleet owners better optimize the use of their trucks.
In order for developers to provide appropriate tools and services and users to effectively justify the significant investment in AI capabilities, identifying and aligning the tools with the intended use cases and applications will be critical.