4 Myths of Industrial AI, Debunked
Artificial intelligence (AI) has permeated nearly every facet of modern life. It recommends the best commute, suggests music or podcasts during the journey, powers countless applications and machines throughout the day, and recommends shows or movies to stream when you get home.
AI is here to stay.
Industrial AI can help manufacturers maximize uptime with equipment monitoring and preventative maintenance programs and identify loss yields and defects (Cavallo) . And its predictive capabilities can create learning and forecast demand models (Koev).
However, AI has struggled to reach widespread adoption in industrial automation use cases. Many companies are still grappling with the basics and are hesitant AI can deliver meaningful returns.
In IBM’s 2022 global AI adoption index report, 34% of survey respondents – about 2,550 businesses from across the world – said a lack of AI expertise is preventing implementation (IBM). Other factors preventing AI adoption included cost (29%), lack of tools/platforms (25%), difficulty and scalability (24%), and data complexity (24%).
Here, we’ll examine those obstacles and dispel common misconceptions about AI in manufacturing and logistics.
#1 Terms are interchangeable and unimportant.
Before exploring AI options, it’s essential to understand the technology's different forms, functions, and feasibility. While some terms may overlap or seem synonymous at first glance, comprehending the nuances of AI is the first step in determining if the technology is the right fit for your needs.
Algorithm: a set of instructions and calculations that help a computer achieve an objective. A “learning” algorithm uses trial-and-error and learn-by-example methodologies to optimize production processes without human intervention.
Artificial intelligence: a group of computing techniques that attempt to mimic human decision-making, using automation to perform tasks that are difficult for humans using image recognition natural language processing, and other technologies.
Deep learning: an AI technology designed to automate complex and highly customized applications. Processing takes place via a graphics processing unit (GPU), which enables quick and efficient analysis of vast image sets to detect subtle defects and differentiate between acceptable and unacceptable anomalies.
Edge Learning: an AI technology designed for ease of use. Processing takes place on-device, or "at the edge," using a pre-trained set of algorithms. The technology is simple to set up, requiring smaller image sets (as few as 5 to 10 images) and shorter training periods than traditional deep learning-based solutions.
Machine learning: Computing processes that can improve outcomes without human programming. Machine learning algorithms train a computer to seek success and avoid failure millions of times to generate learning outcomes.
Machine vision: Rules-based algorithms that identify specific characteristics of an object. Though machine-vision tools work much faster than the human eye, AI can dramatically improve these tools’ accuracy and effectiveness.
#2 AI will replace jobs and foster distrust among employees.
The myth of emerging technology replacing jobs could likely be traced back to the invention of the wheel. The truth is a bit more complicated.Advances in industrial technology, including AI, are rarely conceived in a vacuum. They’re designed to improve performance, efficiency, quality, and capabilities. It’s easy to see why internal combustion and steam engines effectively replaced horses and buggies, or how the telegraph opened new lines of communications compared to hand-delivering letters. These innovations succeeded other forms of technology. Although engines ousted the horse and buggy, the technology created an entirely new industry while enabling mass transportation, transforming logistics, personal conveyance, and shipping.
The same can be said for AI. Instead of AI replacing jobs, companies are discovering that employees can work alongside AI to achieve greater productivity and open new possibilities.
AI can reduce the quantity of mundane, repetitive tasks, empowering workers to address other creative or high-skill functions. In 2018, a New York-based charity began implementing AI for data entry tasks, which contributed to lowering the firm’s annual turnover rate from 42% to 17% (Knight).
The technology is being widely applied to manufacturing and logistics to address the ongoing labor shortage and other chronic issues. When paired with robotics, AI can facilitate tasks such as object avoidance and surface mapping to deliver goods throughout facilities. When coupled with machine vision systems, AI can perform repetitive, albeit essential, quality assurance tasks including part absence/presence detection and inspection (Gow).
Leveraging AI to perform mundane operations enables facilities to reallocate resources toward more intensive tasks and assist front-line workers by offsetting their workload.
#3 Industrial AI requires thousands of images and large datasets
The reality of this misconception can be summarized by one of engineers’ favorite sayings: “it depends.”
AI is a broad field, encompassing numerous types of technology that can be applied in a myriad of ways. For AI to address complex applications such as detecting anomalies on weld seams or analyzing stitching patterns in textiles, the technology must undergo extensive modeling, development, and testing, making data-intensive deep learning-based solutions a suitable candidate.
However, more straightforward forms of AI can address similar tasks, including defect detection and classification/sortation. Edge learning technology, for example, only needs 5-10 images to train and can be deployed by front-line personnel, no experience required.
First, an operator trains the system based on the application. For example, in a part inspection scenario, the user would present the system with images of an acceptable part and parts with defects.
Using only a handful of images, edge learning technology leverages advanced algorithms to differentiate between acceptable and unacceptable parts. Once the system is trained to distinguish good parts from the bad, users can deploy the solution on the manufacturing line.
#4 You need a doctorate and a team of data scientists to implement AI solutions
Developing, designing, and testing AI requires a refined skillset, but using modern AI solutions can be deployed by front-line personnel within minutes.Cognex edge learning solutions run entirely in a smart camera equipped with integrated lighting, an autofocus lens, and a powerful sensor, all of which work together to deliver precise inspection
Since it requires neither specialized knowledge of machine vision nor AI, line engineers can train the technology using their existing knowledge of required tasks. By identifying and clarifying the relevant parts of the image, advanced imaging hardware and edge learning-powered software reduce computational load compared to traditional deep learning approaches.
AI isn't a fad or a specialized technology applicable to specific markets; it's a vast field that can aid the industrial sector in numerous ways. As the technology evolves, it's become more user-friendly. It’s been field-tested in manufacturing and logistics, delivering streamlined quality control, improved product traceability, and enables facilities to identify defects earlier in the production process.
Specialized AI has been used to automate a specific task by analyzing data and patterns to guide future actions (Autor, Mindell, Reynolds). For instance, specialized AI has been applied in manufacturing and logistics operations to inspect parts, confirm the absence or presence of certain components, and sort packages.
Edge learning, in particular, is designed for rapid deployment, requiring only a few images to differentiate between acceptable and unacceptable parts, with all processing taking place on a single device. It's a technology that line engineers can implement in a matter of minutes, assisting operators by streamlining workflows, improving product quality, and boosting efficiency.
Research and information regarding AI manfacturing and procedures described above were sourced from the following:
- Cavallo, Christian. (2022). Machine Learning and AI in Manufacturing. Retrieved from: https://www.thomasnet.com/articles/engineering-consulting/machine-learning-for-manufacturing/
- Koev, Asparuh (2022, September 29). The Top 5 Impacts of Artificial Intelligence (AI) in Logistics. Retrieved from: https://supplychaingamechanger.com/the-top-5-impacts-of-artificial-intelligence-ai-in-logistics/
- IBM (2022, May). The IBM Global AI Adoption Index 2022. Retrieved from: https://www2.deloitte.com/cn/en/pages/consumer-industrial-products/articles/ai-manufacturing-application-survey.html
- Knight. (March 14, 2020). AI Is Coming for Your Most Mind-Numbing Office Tasks. Retrieved from: https://www.wired.com/story/ai-coming-most-mind-numbing-office-tasks/
- Gow. (August 28, 2022). The Labor Shortage is Killing American Manufacturing. Here’s How AI can Bring It Back to Life. Retrieved from: https://www.forbes.com/sites/glenngow/2022/08/28/the-labor-shortage-is-killing-american-manufacturing-heres-how-ai-can-bring-it-back-to-life/?sh=7c6185407374
- Autor, Mindell, Reynolds. (January 31, 2022). Why ‘the future of AI is the future of work’. Retrieved from: https://mitsloan.mit.edu/ideas-made-to-matter/why-future-ai-future-work