Fans of the Battlestar Galactica franchise are familiar with the Cylons, an advanced reptilian race that ended up taking over humans, ultimately creating their own robotic offspring to serve them, run their vast empire and form their military. Later when these robots betrayed their original creators and overran the empire, killing off the master Cylon race, this storyline felt like a cautionary tale of extinction at the hands of technology.
Technology moves so fast now it is not surprising some fear artificial intelligence (AI) developments can run amok. Clearly AI is taking over technology, but arguably in a good way, and is now a part of our everyday lives from our Google maps, to our Amazon accounts. AI is now a vital link in cloud computing, and is on our smart phones and in our new cars, and increasingly, in our many machines and gadgets. It is giving us the kind of assistance we not only crave, but are fast becoming dependent upon.
For manufacturers, AI is changing the nature of business too, with connected machines on the shop floor. Once relegated to science fiction or academia, IBM’s Deep Blue’s success in the famous chess game of 1997 broke through the shroud and showed the public what real artificial intelligence was capable of achieving. (And the trend continues with AI still mastering – and winning– games against humans.)
But the plan for AI’s development was always to solve the problems of business. While others coined the actual term Artificial Intelligence, Edsger Dijkstra, an early programming pioneer is credited in 1956 with ‘inventing’ it. Working on one of the world’s first computers, based at the Netherlands’ Mathematical Center, Edsger believed computer science could solve many problems by algorithms. Dijkstra laid the AI foundation by the use of algorithms for machines (computers) to perform intellectual tasks, and exploded the myth that only human intelligence could solve complex problems or create solutions. When we speak of AI today, it is of this same ability for machines to perform intellectual tasks. These algorithms and their connectivity to the data nodes are at the very foundation of AI, and create a forest of properties and relationships in the data to reduce complex problems to everyday solutions.
AI refers to machines that carry out these complex tasks, and machine learning (ML) is an AI subset where complex computer algorithms are developed to recognize patterns in large volumes of data to solve related problems, independent of human intervention. Like AI and ML, Big Data is becoming an even more popular term, and refers to the overall data size. Like AI and ML, Big Data is seemingly running and connecting every link of the data input in our lives. When we input personal information onto our computers and smart machines, AI is there to quickly learn who we are and update all connected data banks.
A recent survey of sales, service, and marketing professionals in the U.S. shows a growing effort for enterprises to capitalize on AI. Right now, 37% of all corporate teams are using AI, and a further 22% are evaluating the solutions that AI could achieve for their enterprise.
For manufacturers, AI can be used for for the purpose of taking on tasks (and freeing up workers) to mine data. Mining data for analytics and predictions is taking hold in manufacturing. ML has been successful in many areas of manufacturing. It is already known to the public where an encounter with a machine may have assisted in transmitting an image or facial recognition, or given a prediction of a condition or disease based on an electronic health record.
Today, there are three areas of cloud computing where AI is helping manufacturing enterprises: ML algorithms, parallel processing and Big Data. Here is a look at these three areas.
ML really began in earnest in the 1980s when computers were first able to learn. Today, ML (sometimes called deep learning) is the term to refer to these neural networks, or cognitive computing which allows machines to interact naturally with us. ML is the machine’s ability to keep improving its own performance, and for manufacturing enterprises, recent developments show ML is effective.
ML operations are now achieving an almost superhuman performance in a wide range of activities like detecting fraud and diagnosing disease. AI’s ML footprint is now in almost every industry and enterprise from manufacturing, retail, transportation, finance, health care, law, advertising, insurance, entertainment, to education. Perhaps the best example of how business can use data from ML algorithms is to look at Google Maps. Google’s main source of this data comes from our smart phones, but like all navigation applications uses the aforementioned Dijkstra algorithm to find the most efficient route to the requested destination. And Google maps seem to do it best. They are such a mainstay on most smart phones and computers that we no longer stop to think of the successful use of these ML algorithms, or consider its complexity. It is Dijkstra’s algorithm that is the basis of this “shortest path” problem solving route solution and today’s version is insanely fast.
As a bit of history, Dijkstra discovered the AI algorithm when he designed a program back in the 50s that would find the shortest route between two cities in the Netherlands, using an AI-inspired road-map. To the question, “What’s the shortest way to travel from Rotterdam to Groningen?” Dijkstra answered it by designing a map of nodes (64 cities in the Netherlands) for which the algorithms would deliver the shortest path.
Today, a pizza delivery business using Google maps gives the delivery driver’s cell phone the current traffic conditions to show the best route to the customer. Not only does the pizza arrive hotter and faster but the route driver saves time, gasoline, and is back faster, to deliver the next pizza to a another happy customer. A simple app using AI gives instant pay off.
Computers can be amazingly proficient when it comes to performing tasks, but these computations are not always efficient. Many tasks or requests take enormous amounts of time and data. Some of the problems can take microprocessors hours, days, even years to solve. One way to get to a faster solution is to use a powerful processor with nanotechnology. However, this is a very expensive ‘fix.’
Parallel processing employs more than one microprocessor to handle parts of the same task. Parallel processing is an operation or process that splits the task into different parts, and executes them simultaneously on different processors attached to the same computer. By using multiple processors, this method shoulders the load better since they work at the same time to solve the problem. The value of using AI’s parallel processing as a massive problem solver is an easier expenditure for the likes of Google, Intel, and IBM; all have all rolled out neural processors, managed by AI.
Lean flow manufacturing that uses parallel processing is said to increase efficiency by 300%. Lean flow refers to how a modernized, smart machine operation combined with parallel processing will increase profitability. One example where manufacturing needs to quickly react to market demands and change by using parallel processing methods, is in the Sports Nutrition Drink Industry. Sports drinks are an in-demand item, but they are dealing with the complexity of diverse consumer tastes, an overall increase in inventory tracking of SKU (stock keeping unit) numbers, and the reality of many regulations dealing with allergens and hygiene standards.
Sports Nutrition manufacturers find it difficult to cope with increased customer demand and need bigger processing equipment and more space in which to locate it on the shop floor. Most process systems used to produce such high value nutritional products are linear – they start at one end with raw materials and out of the far end of the process comes the final product. For some manufacturers, the same approach, and often the same equipment, is used whether it is a small or large production run of a particular recipe. Problems occur when an ever increasing number of recipe changes further down the process line destroy the effectiveness of conventional processing. Down-time spent in change-overs between recipes not only slows down the process, but imperils operations, particularly where cross-contamination is a real risk. A more modern and efficient way is to apply a parallel processing approach whereby a number of batches of different recipes run through the process line, all at the same time.
Big Data is making a big difference in how AI and ML improve manufacturing. When it comes to big data analysis, the end game is usually to make machines smarter and to expand needed functions. In an article dated July 2014, How Big Data Can Improve Manufacturing, authors Eric Auschitzky, Markus Hammer, and Agesan Rajagopaul effectively make the claim that the era of Big Data is now here. The authors state that for manufacturers, Big Data is a necessary tool for the shop floor in order to improve yield, and aid profitability in which “process complexity, process variability, and capacity restraints are present.”
The article describes how one large European maker of specialty chemicals and detergents boasted a long history of process improvements and its average yield was consistently higher than industry benchmarks, making it a success story that even its own engineers would not challenge or work to improve.
However, unexpected insights emerged when the company used neural-network techniques, and big data analytics, to measure yield. Looking at factors like coolant pressures, temperatures, quantity, and carbon dioxide flow, the data analysis revealed how levels of variability in carbon dioxide flow prompted significant reductions in yield. By resetting the parameters accordingly, the chemical company was able to reduce raw material waste by 20 percent and energy costs by 15 percent, improving overall yield. It is now implementing these advanced process controls – all because the data challenged them to increase yield!
Change is happening daily through AI applications. Consider the DRU (Domino’s Robotic Unit) Assist. This natural language AI chatbot is now available through Domino Pizza’s online ordering app in some markets. It lets people speak to place their order, instead of scrolling and tapping through the app. A text-chat version, launched on the Domino’s website, now allows this pizza chain giant to use a chatbot to take all food orders, enabling workers to focus on what they do best – making pizzas.
AI will not take over our world, or kill our jobs but it will transform our business technology and improve our lives. We have only begun to scratch the surface. AI is beginning to be the vital ingredient in business enterprises, and should lead to better outcomes and more satisfied customers.