Business lessons can come from the most unexpected places. I had a fascinating experience that showed both the strengths and weaknesses of using AI tools. The tools can be used to achieve previously unthinkable things but without a human touch their functionality can be limited.
The Strengths and Weaknesses of AI Translation
This past Friday I was in Michigan visiting my mother. I brought my son so he could spend time with his grandmother, and I could spend time working. On Sunday, her church had a special Unity service with another church in her denomination. The other church is composed primarily of African immigrants, most of whom spoke Swahili rather than English. As a sign of unity in the faith, part of the service was conducted in Swahili with no direct interpretation. It was an interesting experience as I, despite spending much of my life aboard, have rarely heard Swahili.
In the past, listening to part of the service in an unfamiliar language would be an interesting novelty and a little more. This service was different. The church bulletin contained a QR code to Wordly.ai, an AI translation company. Wordly allowed us to select a language that we wanted to receive the interpretation in. My son and I experimented first with Russian, then Spanish, and finally English. The interpretation seemed to be fairly accurate. The generated monologue was reasonable.
Yet the program started to break down as the speakers moved away from readings from the Bible to prayers. The app struggled to follow along and started to produce a mixture of intelligible and unintelligible statements. Look at the results above. One thing you should notice is that the auto-detection believes that the language spoken is Arabic. You may remember earlier that I stated it was Swahili. Swahili is a Bantu language (a different language family than Semitic that Arabic belongs to), yet a significant number of Swahili words are loan words from Arabic.
Will later versions of this translation tool be able to solve this challenge? Maybe. Language is enormously tricky and it will be difficult for AI models to pick up all the various nuances to truly provide the level of understanding required. I am interested in seeing how it goes but for now; it is not the total solution to international communication that we need.
What does this have to do with Manufacturing?
Language is not manufacturing but what they share is that they are human endeavors and thus are rarely as cut and dry as we would hope. AI tools have yet to gain real abilities to discern complex and often subtle differences that all organizations have.
AI tools are great assets for manufacturing operations. We have spoken about AI tools for visual inspection many times in the past. AI integration into Enterprise Resource Planning (ERP) can be a great asset to quickly organize and represent data. Companies are doing great work when it comes to planning machine maintenance and robotics with more advanced learning are further increasing overall efficiency.
These are great tools, but they still need humans to deploy them correctly and discernment to figure out how to apply the data they generate. Understanding the strengths and limitations of each tool at your disposal is key to understanding what tool you need and how to most effectively utilize limited resources.
Expertise and understanding of how each of the processes work together that only humans have is still critical to advanced manufacturing. Are you thinking about implementing more AI tools into your manufacturing operations? What problem are you trying to solve? Let us know below.
What can MTG do to help you improve your operations?
