Harnessing the Third Wave of Artificial Intelligence

Since artificial intelligence (AI) was first coined in 1956, it has been the “next big thing.” Over the years, AI has evolved as a tool for general applications to several vertical applications – from automotive to utilities and energy.
Today, it can be found everywhere, touching individuals’ lives, and changing how industries do business. Now, computers can connect causes and effects from their surroundings and identify patterns and make new suggestions. For example, in manufacturing, computer vision systems in assembly lines can analyze quality defects in real-time with spot-on accuracy. For these and other reasons, we believe that AI is both commercially ready and emerging technology.

AI makes it possible for machines to learn via experience, reason like humans, and sense and build knowledge. Molex Ventures, Molex’s corporate venture arm, identified some of the most common techniques used in AI applications to achieve these capabilities.

• Learn. The most common technique used in many AI applications to learn is machine learning, a set of technologies made up of algorithms that operate like a needle weaving a piece of fabric by slowly stitching multiple pieces together to accelerate a computer’s ability to learn.
• Reason. In AI, the reasoning is essential for computers to derive logical conclusions and make predictions from available data. At Molex Ventures, we believe that, because it makes real-time processing data offline possible, edge computing is a critical enabler of AI.
• Sense. AI enables computers to sense, detect, and use vast data to make intelligent decisions. Sensors detect changes in an environment and provide that information to other electronics. AI needs accurate sensors to provide data streams required for processing insights.

The past, present, and future of AI encompasses “three waves,” as described by the Defense Advanced Research Projects Agency(DARPA). In the first wave, rule-based systems were developed; the second wave of statistical learning is where we are today. Each wave represents capabilities that can unlock more commercial use cases and applications.

At Molex, we’re getting ready for AI’s third wave, which will be a world in which computers have the potential for causal understanding and contextual reasoning. In other words, they will be able to connect causes and effects from their surroundings. Not only will computers be able to identify patterns, but they will also explain decisions and make new suggestions.
The future of AI-embedded products at Molex is nearer than you think. When it comes to new product development, we expect AI software to recommend new product features based on performance and customer preference data. In production, quality systems will help us predict product defects and recommend countermeasures that we can take in product design and manufacturing before developing final designs.
There are no AI models that can effectively take data, physics-based models (i.e., models that describe the physical world in classical mechanics), and human judgment and integrate them into a functioning AI system. Yet, over the next few years, we anticipate software availability for new learning methods that will give us previously unimaginable capabilities. In short, by harnessing the third wave of AI, we will have a greater understanding of how to solve our industry’s—and our customer’s—most daunting challenges.

• Generalized software platforms like Google and Amazon will become commodity platform providers for agnostic industry vertical services. Companies offering industry agnostic solutions are too late.
• AI models designed to solve industry-specific problems will become a new source of AI innovation. Simultaneously, architectures for platform serviceability, the ability to accommodate different devices, and the ease of scaling out architectures will create value.
• Population health management platforms, which are proliferating, could be essential for driving healthcare efficiency and accuracy. Platforms offering patient scheduling and back-office will gain the most traction at first, followed by diagnostic software capability. AI will accelerate point of care devices featuring diagnostics that receive a diagnosis without visiting a doctor.
• AI will augment manufacturing capabilities to ultimately create a fully automated facility. New quality systems will automate product traceability and provide countermeasures. Demand systems will match supply with real-time pricing and capacity planning. Advancements in computer-animated design (CAD) and printed circuit board (PCB) design will reduce engineering time and product to improve reliability.