Following the launch of DeepHow last month, media coverage continues to roll in. Many thanks to Automation World’s David Miller for his story in the Factory/Workforce section of the magazine.

How Artificial Intelligence is Being Used to Streamline Training

DeepHow is using artificial intelligence to make the creation, delivery, and consumption of digital training materials more efficient.

Maintaining a skilled workforce is vital to the success of Industry 4.0. Despite the fact that continuously updated skills training is a well-recognized facet of the digital transformation, many industry observers see the current situation as dire.

The Manufacturing Institute and Deloitte estimate that while 4.6 million manufacturing jobs are forecast to be created over the next decade, as many as 2.4 million of those positions are likely to remain unfilled due to the continuing skills gap. This issue is exacerbated by a wave of retirements on the one hand, and a lack of enthusiasm for manufacturing among young people on the other. Typically, conversations aimed at addressing the problem have been centered around the formation of apprenticeship programs, better STEM education at the primary and secondary levels, and means by which more one-on-one mentoring between experienced and newer employees might be facilitated in plant environments.

Though much progress has been made with these approaches, a number of challenges remain. First of all, while many are hopeful that social distancing requirements will soon be loosened, the past year has rendered hands-on peer-to-peer instruction particularly difficult. Moreover, educational programs aimed at helping new employees acquire the skills they need to enter the manufacturing field aren’t always as effective in addressing the need for the current workforce to continually update its skills, as digital transformation often brings rapid alterations to prevailing work processes.

Fortunately, much of the same technology facilitating Industrial Internet of Things (IIoT) data interchange can also be used to foster digital knowledge transfer by enabling workers to access unified multimedia databases of training materials and other content from any location. This can take many forms; for instance, robotics supplier Universal Robots offers web-based training courses through its UR Academy platform, augmented reality (AR) headsets are increasingly being used to offer real-time remote assistance to field service professionals, and the ubiquitous presence of mobile devices such as smartphones is increasingly granting workers an always-on portal to any information they might require.

DeepHow is an example of a company employing the latter approach. Its software allows end-users—who in this case take on the role of content creators—to record how-to videos that instruct workers on performing specific tasks and processes. From here, the content is uploaded to a secure knowledge database accessible to employees. While these core capabilities may seem relatively rudimentary, they are complimented by a suite of AI-powered tools that further streamline the curation and distribution of training materials.

This is how the technology works: First, DeepHow automatically transcribes the spoken dialogue in the video, providing searchable, time-stamped text to accompany the video. In addition, the application is capable of translating the text into numerous languages that can be appended to the video as subtitles or read as plain text.

Then, DeepHow analyzes the recorded content to break it down into a series of chapters or steps, allowing viewers to more easily navigate to the section(s) that address their needs. The content creator can also alter the arrangement of these segments via click-and-drag mechanisms. They can also retitle them if they are not fully satisfied with the results the AI provides.

Once the steps are established, each one becomes its own module so that if small parts of a process change or new protocols are implemented, a single portion of the overall training video can be replaced without needing to re-record the entire series of steps. Finally, other details and information can be attached to individual steps, such as additional text, graphs, schematics, and other images.

The company estimates that its smart video delivery system is ten times faster than traditional video editing and dissemination techniques and results in a 25% improvement in worker performance. DeepHow has recently completed several months of piloting and is now targeting manufacturers and field service operations for further deployment.