It’s difficult to identify exactly what Stephen Hawking will be most remembered for. Is it his contributions to cosmology and our understanding of space-time? Or maybe quantum mechanics and the very key to unlocking reality? Maybe it’s the fortitude of the human spirit and mind, which he proved can thrive even as the body atrophies.
Hawking had ALS for 55 of his 76 years, a disease which stripped him of his mobility in the late ’60s and of his ability to talk in 1985. The theoretical physicist and professor relied on a computer to speak for him, slowly translating typed words into a tinny, mechanical voice. As Hawking lost more of his motor functions, his text-to-speech system evolved with word-prediction algorithms and controls based on facial movements—and later brain waves. This allowed arguably the greatest mind in the world to more quickly and efficiently express his thoughts.
It took many brilliant engineers and decades of computing advances to meld mind and machine in order to help Hawking overcome the limitations biology set upon him.
|Stephen Hawking: 1942-2018|
Considering that a manufacturing business operates much like a living organism, with the CEO as the brain, we must accept that limitations exist here as well. And like Hawking, manufacturers must learn to rely on intelligent machines.
For Hawking, this was to answer questions and ask many new ones concerning the infinite expanse of the universe. For manufacturers, partnering with artificial intelligence will allow CEOs to ask the right questions, and using the millions of data points streaming in to provide solutions.
“AI changes everything: business models, operational models, how work gets done, how workers are trained,” says Cliff Justice, a partner at KPMG who leads the firm’s cognitive technology, artificial intelligence and automation investments. “It’s a transformative field to say the least.”
For the purposes of the article, the field of AI encompasses machine learning, or the computer programs that learn and improve on their own via algorithms. At some point you’ll need at least cursory understanding of how this all happens, but not today. Right now, you need to know that it reconciles godlike ambitions with primate brains. You want omniscience concerning equipment, employees, suppliers, customers? There’s probably one or more AI solutions that can provide it. If not, your team can build it.
Augmenting, Not Replacing
Justice’s team at KPMG used IBM’s AI platform, Watson, to audit massive loan documents to ensure the bank graded a commercial loan correctly before it was bundled into a security. By themselves, humans would only read a sample the documentation, but a computer can process volumes in an instant.
“We don’t turn it loose on its own,” Justice adds. “It’s augmenting, not automating our people.”
He says the program provides prompts for the auditor, just like how online tax software would open a pop-up box if it detects a problem or needs more info.
“It’s not about saving time, it’s all about improving accuracy and quality beyond what a human could possibly do,” Justice explains.
Manufacturers are often content with incremental improvements in time, accuracy and quality. As long as you’re “continuously improving,” you’re on the path for success. However, due to AI, in ten years the current incremental changes catalyzed by kaizen and lean manufacturing might look like absolutely pedestrian. Heck, this could happen just over the next year.
In 2016, GE boasted that its Predix platform could increase an industry’s performance by 1%. Last year across GE’s Brilliant Factories, Predix yielded much greater improvements. In India, equipment effectiveness reportedly increased by 18%, while a Michigan plant cut downtime by as to 20% by applying IoT sensors to monitor wear.
Granted, GE has a great head start and few should expect to match this pace, but many companies aren’t even at the starting line in this race to embrace AI and its benefits. Only 4% of chief information officers globally have implemented AI, while 46% are planning on it, revealed Gartner’s 2018 CIO Agenda Survey. In the short term, only 25% have plans for this year.
For companies with $50 million revenue—and higher investment capital, the stats are better. According to tech research firm Vanson Bourne, 80% of these enterprises have deployed some form of AI, such as machine or deep learning. Less than half characterize its use to be significant and deployed operationally. In the U.S., 61% have “lots of room for further implementation” or plan to deploy in the next two years.
What all this tells us is that manufacturing isn’t about to change in a profound way; it already has.
Fourth Time’s a Charm
The dawn is just breaking in this fourth industrial revolution. Brand new dreams and ambitions crown the horizon: skylines of smart factories, smart cities, smart Martian colonies. It feels like the future of old, in fact. And where we’re going, we’ll still need roads, but not drivers.
But we’re not quite anywhere yet. Manufacturing CEOs will need to harness the power of AI to get there, and will of course present some challenges. AI is as broad a term as “material handling.” And you can’t go to Grainger and say, “I’d like two AIs, please.” Google’s open-source Magenta can compose music, while IBM’s Watson is famous for defeating Jeopardy! champ Ken Jennings. These are very siloed use cases, and while novel, won’t help manufacturing.
The process really starts with a simple question: What is the problem you need AI to solve? Energy efficiency? Equipment maintenance? A simple chat bot to expedite customer service?
“AI is an applied science,” says Gene Chao, Global Vice President IBM Automation. “If there’s no application of it; it’s just a cool thing.”
One use Chao points out is extracting invoice data across different formats. He refers to a program that pulls 80 to 90% of data, including vendors, dates and currency, which can dramatically reduce processing time.
Another clear use case is leveraging data from IoT sensors to predict when critical machinery will fail or need to be serviced.
|IBM Cognitive Visual Inspection solution enables manufacturers to improve productivity of their manufacturing and assembly processes while reducing operations costs and improving product quality. Working side by side with human inspectors, the more the solution inspects, the more it learns.|
Data, of course, is the key to making good decisions. And it’s everywhere. The temperature and vibration sensors on machinery, GPS data on trucks and speeds of AGVs in the warehouse. But it’s relatively inert without a human to assign value to this data, Chao says. An AI program merely sees numbers. It needs a person to define how important those numbers are and decide what the course of action should be.
In this new era of engagement, reasoning and judgment are at the forefront.
For the C-suite, this has never been easier. As the vertical integration of machine learning continues, ERPs and CRMs, the front and backend, it’s all connected, allowing decision makers to recognize patterns and trends they can act on.
“The walls of those domains are transparent today, but someday there won’t be walls,” Chao says.
Rise of the Machines
IBM’s Maximo Asset Management is one system already tearing down those walls. By allowing users to sense, communicate with, and diagnose problems on connected devices and machinery in the plant, the company says it can help reduce unplanned downtime by up to 47%, while the data gathered from workflow processes, throughput and yield can reduce defect rates by up to 48%.
Currently, controls and automation leader, ABB, uses IBM Watson in its wind business to predict output based on wind speed forecast.
“This helps the operators commit to the production of power in the energy markets,” says Guido Jouret, ABB’s chief digital officer. “It also helps with optimizing the best time to perform maintenance.”
Jouret says ABB also uses Watson to analyze customer issues and prescribe solutions, as well as identifying upsell/cross-sell opportunities. At its Heidelberg plant, ABB also employs adaptive algorithms that learn from previous quality control efforts to increase the precision of future tests.
“The big benefit in the industrial area is that a few percentage points of improvement are typically carried forward over many years because industrial equipment can last for decades,” Jouret says, “so the benefits can be substantial.”
An operator can also use machine learning to train collaborative robot. After manually moving the robot’s hand to the locations it should in a pick-and-place task, it will be able to mimic the motion. “This is much quicker than writing code to configure the robot,” Jouret says. “As we add more smart sensors, robots can operate in more collaborative fashion, which makes it possible to use more of them.”
Prescription for Success
Trendforce estimates that last year the global smart manufacturing market was more than $200 billion and will expand by 60% to $320 billion by 2020.
Downtime is not the only worry if a part goes bad. A European pharmaceutical manufacturer found that integrating Oracle’s IoT platform into its process can reduce waste.
“Once in while valve goes bad because of wear and tear of from the chemicals and ruins entire batch; that costs tens of thousands of dollars,” explains Atul Mahamuni, Oracle’s IoT Cloud VP. But they have integrated prescriptive maintenance, empowering the machine to make real-time decisions based on the data, like that a valve is about to fail.
“The moment that happens, the system can detect it and stop production and you are not going to have any wastage,” he says.
Oracle takes machine learning to a new level by allowing the AI to decide which algorithm is best suited for a certain process. While this hints at a future where machines make unwanted decisions, Mahamuni says in this case it’s more like an electrician choosing the right pair of custom pliers for a specific job.
“The user still trains the model and selects the right set of parameters and tunes the predicative model,” he says.
With all of this early evidence in, it would seem that manufacturers clearly stand to reap new benefits from deploying AI successfully. But who will be around to celebrate the factory’s historic success if everything is automated?
There is a dark side to AI, in case you missed every science-fiction movie ever. In 2016, Hawking, at the launch of Leverhulme Centre for the Future of Intelligence, noted the dichotomy:
The potential benefits of creating intelligence are huge. We cannot predict what we might achieve, when our own minds are amplified by AI… In short, success in creating AI could be the biggest event in the history of our civilization. But it could also be the last–unless we learn how to avoid the risks. Alongside the benefits, AI will also bring dangers like powerful autonomous weapons or new ways for the few to oppress the many.
Aside from the automation anxiety stirred up by science fiction movies dating back to Metropolis in 1927, there are concerns over employment and safety. An infamous 2013 Oxford study concluded 47% of U.S. jobs could be automated by 2038.
The problems AI poses should be balanced against their benefits and what humans can control, says Cliff Justice. The investor doesn’t put much stock in one Oxford study.
“That study is interesting in a way, and in another way it’s a little bit of fear mongering,” he says. “No time in history of humanity has technology led to a lower standard of living.”
KPMG predicts 5 million new jobs will be created because of intelligent automation.
“This is like electricity was 100 years ago,” Justice says. “The cloud is making AI accessible to so many inventors, and for pennies per API.”
Following this monkey/typewriter logic, innovators increase, as does the amount of must-have innovations. This spurs a need for more companies to make or support said innovation, which will need people to work at them.
|The infinite monkey theorem in effect.|
And AI for the foreseeable future is not sentient and needs human oversight, so factory jobs may change, but won’t go away.
“As you automate repetitive, cognitive work in a factory or white collar setting, you’re freeing up talent, you’re dropping the cost for lower value repetitive activities that do not generate profit,” Justice says.
To stay competitive, he assumes most companies will invest these profits in engaging customers and expanding their base, creating higher skilled jobs. These jobs will be focused on asking the right questions and identifying the big issues. More data coming in will need more human brains to figure out how to use it.
Justice sums this up perfectly: “Should you adopt AI? You’re going to have to. You’ll be irrelevant if you don’t. You might put your company at risk if you don’t. The horse is sort of out of the barn on this.”