Empowering Manufacturing with GenAI: A New Frontier of Possibilities

Imagine holding the key to unlock a technological revolution in the manufacturing world, akin to the impact of the PLC. Generative Artificial Intelligence (AI) is reshaping the manufacturing landscape, with the potential to transform major value streams within the industry. This blog post serves as an introduction to a Research Spotlight on Generative AI, which delves into the capabilities and applications of this cutting-edge technology. Read on, and we'll try to whet your appetite by showing you some examples of what generative AI can do.

The State of Generative AI

Generative AI, including Large Language Models like OpenAI's ChatGPT, has seen rapid adoption in various industries, outpacing social media platforms like TikTok and Instagram regarding active user adoption.

Generative AI comes in various forms, from text generation to image, music, and video generation. Manufacturers have started tapping into its capabilities to accelerate processes, make more informed decisions, and optimize operations. From code generation and troubleshooting in engineering to improved supply chain execution and enhanced customer service, Generative AI is already making waves.

Don't know Generative AI yet? Then, "Try It!"

To truly get a feeling for what a new technology can do, you sometimes have to try it. For most technologies, this is prohibitively expensive and time-consuming; luckily, Generative AI is very accessible. LNS Research will typically ask you to start with defining the business problem; in this blog post, we will do the opposite and start with the technology. Please don't make this a habit.

Here are some exercises that will give you some insight into the capabilities of ChatGPT and other Large Language Models. We encourage you to follow along and try them yourselves. To get started, visit chat.openai.com, bard.google.com, or one of the other suppliers of Large Language Models. We have used ChatGPT-4 from OpenAI in all our examples.

Summarize Business Information for the Target Audience

For the first exercise, ask your favorite Generative AI tool: "Summarize Tesla's business strategy." Here is what the dialog looks like:

You will get a nice summary of the primary strategies Tesla is using. This is both impressive and valuable. However, this response may not be relevant to the intended audience. To make it relevant, create a second prompt: "Explain the strategy to a process engineer." You will now see a response tailored to a process engineer's skillset.

Not only did ChatGPT-4 summarize a very complex topic, but it also communicated it in a language that suited the target audience. Many use cases are similar to this; we challenge you to think about what those use cases could be in your industry.

Experience the Hallucinations of a Generative Pre-trained Transformer

Let's stop thinking about business strategies for a moment and have some fun. Open a new tab in your web browser with ChatGPT in it and paste the following text into the prompt: "It was a warm summer day. We were sitting under a tree and looking out over the green field."

Observe how ChatGPT responds; it will continue to write an essay based on your opening words. You did not even ask it a question; it just strung the words together to create a new hallucination with no roots in reality and with the same level of confidence as the fact-based Tesla strategy. You can experience the probabilistic behavior of ChatGPT by asking the same question again and experiencing a different result.

The scary part is that ChatGPT did nothing to warn you about its hallucinations. This is one of the many areas where you have to be careful. While Generative AI can be helpful, it can also lead you astray.

Let's get technical: Write a bit of Code and do some Analysis

Finally, let's ask the model to generate code to solve/analyze stock market data using the following prompt: "Using Python, write the code to connect to a source of the S&P 500 index and analyze the results. Include charts."

ChatGPT is choosing yfinance from Yahoo as the data source and explaining the code step by step.

If you want to run it, you need to install Python and other developer tools (it is beyond the scope of this exercise). Here is what the results look like when running in Microsoft VS Code with Jupyter Notebook:

You may argue that a "20-"and 50-day simple moving average is not a good analysis tool, but that is an example of a vaguely defined prompt; try for yourself and see if you can do better. If you are like most businesspeople, then you see where this technology could be applied to other, more relevant areas of analysis.

ChatGPT excels at the code writing task because the training set is well-defined and complete. Programming languages have a clearly defined syntax that is required for the code to work, and the internet is rich with examples about how to solve a myriad of problems. However, you will find that even for a task with a well-defined training set, ChatGPT is not perfect, and you must review any code that you use.

The Opportunities and Challenges Ahead

"You can think of Generative AI as the end of your beginning. In essence, it's the
starting point for a much richer journey using a combination of AI technologies," says Ron Norris, Director of Operations Innovation at Georgia-Pacific, LLC (GP). Ron and Mike Carroll (VP of Innovation) have established GP as a thought leader in AI for manufacturing and supply chains. GP has moved beyond using Generative AI by itself and has proved that the combination of Generative AI and Causal AI can optimize and guide complex business processes like procurement and touchless orders.

Using Generative AI is not all smooth sailing, and you will likely fail if you don't have the proper guidance. Many technical, organizational, cultural, ethical, and legal challenges must be addressed to be successful.

Read the research spotlight to gain insight into business alignment, use cases, value streams, pitfalls, and how to navigate the process.