CEO of Neurala, a deep learning neural network software company, and founding director of the Neuromorphics Lab at Boston University.
getty
In the summer of 2012, Google made a big media splash when it showed that its researchers "trained a network of 1,000 computers wired up like a brain to recognize cats." While AI, neural networks and its most recent rebranding, "deep learning," were already established fields with decades of research and countless real-world applications behind them, the world at large (and all its cats) took notice.
Deep learning, a branch of AI that closely mimics how neurons wire and fire, was becoming more powerful: The massive amounts of digital data and compute power needed for training these systems were now available to companies like Google.
Since 2012, applications of AI have expanded to both the consumer and enterprise realms. For instance, AI can be applied to make smart phone pictures more beautiful, delete spam messages, recognize faces, translate languages, make video games more appealing and optimize sales engagements, among many others.
The applications above are all examples of a digital asset (e.g., a raw picture taken on a smart phone, or a paragraph in Italian) being enhanced by a deep learning model to produce another digital asset (e.g., a professionally rendered picture or a clean translation from Italian to English). These can be thought of as "digital-to-digital" applications of AI. Namely, AI uses data to make more useful, meaningful or, occasionally, more beautiful data.
While these are worthy causes for AI, should all our time, effort and money be spent on preventing email spam, cyberattacks and translating foreign languages? There are many more applications where AI spills over to the physical world that affect us more closely or, in a sense, physically.
These "digital-to-physical" applications of AI are more challenging in that they need to migrate from more constrained, less messy environments—where the problem domain is well characterized—to real-world environments that are less structured, forecastable and cooperative.
How challenging is it to break this digital barrier and have AI spill over to successfully interact with a physical system? Suffice it to say that, today, AI can beat humans at games like Chess, Go, Jeopardy and even video games—all of which have environments and actions that can be complex but are also well-defined and limited. Nevertheless, AI-powered machines have a hard time opening a simple, real-world door, where handles come in all sorts of shapes, forms and positions. The real world is beautiful—for AI, it is a mess.
At Neurala, we learned this many years ago while working with NASA. We developed artificial brains for Mars rover-like robots to enable the rover to drive completely autonomously. But, when we think of AI applications that spill over to the physical world today, we think of delivery robots, drones and self-driving cars.
However, there are many other very real applications that are less futuristic, but still important to our everyday life. As the world copes with various states of emergency, manufacturing—one of the most physical industries— is in crisis. With the global supply chain and factories struggling, the pressure is on to get operations caught up so that consumers have full access to the products they need.
In the manufacturing industry, the demand for new technologies—and AI in particular—has skyrocketed. Industrial manufacturers are rushing to implement Industry 4.0 initiatives, such as AI, on the production line. One example of a physical function demanding AI is quality inspections. Pre-pandemic, the task of quality inspections was traditionally performed by human workers. At the time, the labor shortage in the industry meant working double time to get the job done. These efforts were complicated even further with the onset of the pandemic. New social distancing requirements restricted the number of workers who could be on the factory floor at one time. So, how big of a feat would this be for AI? Some estimate that quality inspections are being performed by nearly 600,000 people in the U.S.
By crossing the digital/physical barrier and implementing AI-powered visual quality inspections, the industry can mitigate the crisis and labor shortage. The use of AI removes the barriers that typically slow technology adoption in that it is cost effective, easy to integrate and doesn't need specially trained staff to operate.
AI-based visual inspections are used today to inspect for defects in metal engine parts, check integrity of rugs/carpet, assess whether raw material (such as meat) has foreign contaminants (e.g., plastic particles), check plastic food trays for the right item, inspect quality of baked goods (e.g., bread), determine integrity of vaccine vials and more. These are all real-world, often mission-critical applications of AI technology in challenging physical settings.
The value of digital-to-physical applications of AI is clear, as well as how they can be applied in the manufacturing industry—so what's next? For anyone looking to implement AI across their organization, the next steps are simple. First, you need to take a look at your specific workflows and determine what processes could benefit from AI: Is it a quality inspection, is it predictive maintenance or is it something else? From there, you should partner with a local expert to identify the best solution for the task, so that they can help you integrate it into your operation.
With all its challenges, 2021 will be a pivotal moment for AI: The urgency many manufacturing processes face need real-world, innovative technology to help cope with pandemic disruptions.
Even though its first large-scale incarnations were concerned with spotting cats, the biggest return on the AI investment will affect the very physical objects we use every day: from food, clothes, furniture, all the way to the actual screen of the device we are reading this article on. AI is breaking the digital barrier to make our world a better place, but this time, physically speaking.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
