When Tesla Motors CEO Elon Musk proclaims that artificial intelligence is “our biggest existential threat,” it makes headlines worldwide. But what goes unreported is that the very search engines people used to find Musk’s comments are themselves an example of how AI has subtly but forcefully become a part of everyday, real-world life. When it comes to a discussion of AI, it helps to have a sense of history—as well as a sense of humor.
Thanks to premonitory proclamations by Musk, Microsoft’s Bill Gates, Cambridge’s Stephen Hawking and other prominent technologists, AI has become a popular topic again, after a 20-year cooling-off period. It’s tempting to assume that the “dire warnings” about AI being a threat to mankind were mostly tongue-in-cheek, but the end result is that just as it happened in the 1980s and ‘90s, the hype over AI is again outpacing the reality (virtual and otherwise). The first question that needs to be answered though is: Whatever happened to AI and why did it go underground for so many years?
The answer is, AI didn’t vanish at all, but rather than emerging as a fully-formed incarnation of “The Jetsons” or “The Terminator,” AI went mainstream. People interact with AI every day, in so many ways that nobody thinks of it as the application of computer intelligence; it’s just another app, like Apple’s Siri or speech recognition capabilities in voice-mail applications like Microsoft Outlook. In the industrial arena, supply chain planning solutions are AI-based, and most predictive analytics tools have some form of AI component. (There are numerous types of industrial robots, such as automated guided vehicles, that also have embedded intelligence, but for the purposes of this article, we’ll omit robotics from the discussion.)
The prototypical AI-based application is the expert system, which is basically a solution that embodies the knowledge of a human expert, and can make business decisions through a rule-based methodology that taps into that knowledge to mimic the expertise of the human expert. Expert systems are thus able to extrapolate solutions to unforeseen problems, giving them in effect the capacity to learn. These types of solutions date back to the 1970s, but really came of age during the Reagan years (parenthetically, I covered the AI field for about a dozen years, from the late 1980s until the Internet bubble burst in the early 2000s). And just about every manufacturing sector has developed and deployed expert systems down through the years—demand forecasting for computer production, chemical process control and monitoring, design optimization for automotive manufacturing, defect detection in electronics components, and a cost estimator for piece-part manufacturing, just to name a few.
Always On, Always Available
Lest you think expert systems have faded away, they continue to proliferate (though getting far less attention these days, as the popular media is more captivated these days by high-profile stunts, such as IBM’s chess- and Jeopardy-playing computers).
For instance, as one of the world’s major manufacturers of forklifts, Toyota Material Handling USA (TMHU) oversees support of an installed base of 500,000 vehicles throughout North America. Its dealer network includes 3,000 technicians, who receive more than 18,000 service calls every year. When you consider that roughly 70% of all manufactured goods are transported on a truck, and that the vast majority of those goods have to be loaded and unloaded with a lift truck, the supply chain can slow down or even come to a halt if a forklift unexpectedly has to be repaired or serviced.
TMHU teamed up with Exsys Inc., one of the longest-tenured developers of AI-based solutions, to develop TED (Toyota Electronic Diagnosis), an online expert system designed to speed up the diagnosis and repair of forklifts. A technician enters the appropriate model number and error code into the system, and then TED provides guidance in how to repair the problem, offering photos, diagrams and schematics.
“Now Toyota dealers and technicians have a one-stop source to assist their customers,” explains Cary Howie, TMHU technical services manager. “Less experienced technicians benefit from the experience of technicians with many years of diagnostic knowledge. Having TED on staff is like having another person available 24/7. TED never gets sick, and is always available.”
According to Howie, the cost to develop TED will be returned in the first year of use, with a higher ROI expected as future enhancements and forklift models are added to the system.
Knowing What to Do without Being Told
Predictive analytics is one of the disruptive technologies that will have a dramatic impact on supply chains and the people who run them, according to a recent survey of 900 U.S. supply chain executives conducted by Deloitte Consulting and MHI, a trade association for the material handling and logistics industry. Almost half (44%) of the respondents say that predictive analytics provide a competitive supply chain advantage, especially in today’s era of the “always-on” supply chains, notes Scott Sopher, principal at Deloitte Consulting.
Predictive analytics refers to the ability of computers to crunch through enormous amounts of data (Big Data) and make sense of it all. As the Deloitte/MHI report explains, when applied to supply chain problems, predictive analytics “allows managers to manage inventory better, plan more reliable transportation networks and reduce variability in lead times. This can enhance service levels, lower costs and improve the bottom line.”
One of the better known providers of analytics software is SAS, which describes its solutions as being based on “deep learning” neural networks, a type of AI solution that, in effect, teaches itself how to look for answers without being told exactly where to look. Food giant Nestlé, for instance, uses an SAS demand planning solution to improve forecasting accuracy while minimizing inventory overstocks. The company has roughly 10,000 SKUs to manage, which requires tight synchronization with seasonal and promotional marketing campaigns. The challenge for any consumer goods manufacturer, and particularly one as big as Nestlé, is to maintain high customer service levels without tying up too much capital in safety stock.
Predictability of demand for any given product is highly dependent on that product’s demand volatility, explains Marcel Baumgartner, Nestlé’s global lead for demand planning performance and statistical forecasting. “Especially for products that display wide fluctuations in demand, the choice and combination of methods is very important.”
By applying analytics, specifically a forecast value added methodology, when demand planning for highly volatile products with high volume, Nestlé has been able to more accurately anticipate customer behavior by integrating the impact of promotions and special offers into the statistical models, Baumgartner notes.
Seeing the World through Augmented Eyes
Virtual reality, another branch of the AI field, continues to fire the public’s imagination, but VR has also long had an identity crisis: The coolness factor that comes from high-tech gadgets like head-mounted displays and smart glasses is also largely to blame for the technology failing to catch on in a big way in business settings—it’s just plain hard to take somebody seriously when you can’t even look them in the eye. It’s been well over 20 years since VR first emerged from computer labs into the marketplace, but for all of the buzz that’s swirled around it, VR is still predominantly a solution in search of a problem. While the graphics have improved exponentially since the early 1990s, that improvement has mostly resulted in flashier videogames.
That situation may finally be changing, though, thanks to the latest wrinkle on VR known as augmented reality. The basic difference between AR and VR is that virtual reality involves the simulation of a fictitious world, whereas augmented reality is designed to supplement a real environment (for instance, a manufacturing plant or a warehouse) by offering information to one of the senses (usually vision) that’s not apparent without the AR device. Google Glass is one of the best-known of these types of devices, though augmented capabilities have been in use for as long as VR has been around, and if anything, is an easier sell to business users since the technology is far less disruptive to normal work behavior.
Airplane manufacturer Airbus, for instance, uses what it calls a “connected” glasses tool on the assembly line of its A330 aircraft. These devices, worn by technicians, are used during the cabin installation marking process. This stage in the final assembly involves the positioning of exactly where seats and other cabin furnishings will be affixed in the plane.
The smart glasses, developed by Vuzix Corp., feature a barcode-scanning camera that allows the user to see specific cabin plans and schematics, as well as the marking zone. Once the mark has been made, the system checks the location to validate the process. The user interacts with the augmented reality through voice recognition, another AI technology.
“With our new tool, time spent per aircraft on marking operations is divided by six with an error rate reduced to zero, regardless of the user’s experience,” explains Benoit Rollin, Airbus’ head of manufacturing engineering for A330 cabin furnishing. “Even newcomers, after a short training session, can now be entrusted with this activity.” This is a particularly important consideration for Airbus as the company is adjusting its staffing in preparation for production ramp-ups across its product line of commercial aircraft.
Following a quick development after the project’s launch in January, the first prototype of the tool was available in February for initial testing and validation. To date, the glasses have been evaluated on five aircraft, allowing Airbus to make significant observations of its application on the A330 final assembly line.
Meanwhile, construction equipment manufacturer Caterpillar, which has been using VR in various applications for over 20 years, is currently testing a proof-of-concept augmented reality application that would help field service engineers perform machine maintenance and safety checks. The solution works with smart glasses, but can also be accessed through a smartphone or tablet. One key feature of the solution is that the user can take a picture of every step throughout the maintenance process, and learn from the application if the procedures were performed correctly.
“If you put a hose back in the right spot but didn’t route it correctly, the camera technology would be able to recognize it wasn’t routed correctly and tell you that you did it wrong,” points out Larry Johnson, a field service engineer at Caterpillar. “I think we’re just scratching the surface of what we can do [with augmented reality].”