J U L Y 2 0 2 3 • M A T E R I A L S E V A L U A T I O N 29 2307 ME July dup.indd 29 6/19/23 3:41 PM

which is roughly one quarter the size of the Library of Congress. Second, these models also operate differently than models of the past. ChatGPT and other LLMs are trying to produce a reasonable continuation of the text that was initially provided. Some people refer to this as a “next word guessing engine.” Based on input phrases, these LLMs are essentially evaluating the probability of what the next word should be. What is interesting is the latest generation of models don’t always choose the highest-ranked word. Because some randomness is involved in the model response, these AI chatbots appear to produce much more variety and inter- esting content. For more details on how ChatGPT works, see Wolfram (2023). Potential Benefits Let’s ask ChatGPT about how it might help NDT professionals with their job. (Note: my prompts are in blue and ChatGPT’s responses are in red.) ChatGPT Q&A Part I What are the benefits or positives of using emerging AI chatbots like ChatGPT with work in nondestruc- tive testing? There are several potential benefits to using emerging AI chatbots like ChatGPT in nondestruc- tive testing (NDT) applications. Some of these benefits include: 1. Improved speed and efficiency: AI chatbots can perform tasks much faster than humans, and can work around the clock without breaks. This means that NDT tasks can be completed more quickly and efficiently, which can lead to cost savings and improved productivity. 2. Increased accuracy: AI chatbots can be trained to perform NDT tasks with a high degree of accuracy, which can reduce the risk of errors and improve the reliability of test results. 3. Enhanced safety: NDT tasks can often be dangerous or involve exposure to hazardous materials. By using AI chatbots to perform these tasks, human workers can be kept out of harm’s way, reducing the risk of accidents and injuries. 4. Scalability: AI chatbots can be easily scaled up or down depending on the needs of the project, making it easier to handle large volumes of work or adjust to changes in demand. 5. Better data analysis: AI chatbots can process large amounts of data quickly and accurately, allowing for more detailed analysis of test results and better insights into the health of the materials being tested. Overall, the use of emerging AI chatbots like ChatGPT has the potential to significantly improve NDT processes, making them faster, more accurate, and safer for workers. This reads pretty well. I must admit I’m skeptical of some of these points, but I do feel there are clear opportunities for AI chatbots to assist humans to be more efficient with some laborious tasks. Jill Ross, the ASNT Director of Publications, wrote about the potential impact of ChatGPT on editing in the February 2023 issue of Materials Evaluation, using ChatGPT: “AI has the potential to revolutionize the editing and publishing industry by automating various tasks that are traditionally done by human editors and publishers” (Ross 2023). For example, one application is writing sum- maries and abstracts. AI can take a longer piece of writing and summarize it into a shorter version in seconds, which can be particularly useful for editors and publishers who need to quickly grasp the main ideas of a piece of writing. Another one of the most promising aspects of the latest AI agents is improved language editing for non-English speakers (Kim 2023). Such tools, currently free to everyone around the world, can produce gener- ally good-quality, edited paragraphs in seconds. Coupling this feature with existing AI-based text translation tools, the quality and accessibility of published materials is being positively impacted throughout the world. Taking this one step further, Ripi Singh has been one of the leading early investigators of the potential of collaborative writing with AI agents. In 2021, Singh and Garg created their own interface script to engage with GPT-3 and collaborated with it to research human factors in decision-making for NDE 4.0. The result was published in the Journal of Nondestructive Evaluation (Singh et al. 2021) and was the first NDT paper having an AI agent as a co-author. The article and a more recent blog post (Vrana et al. 2023) provide evidence of the oppor- tunities for practical human-machine co-working at the linguistic and cognitive levels. (Also, see the NDE Outlook article in this issue on page 17 [Singh and Garg 2023].) Beyond just technical writing, AI agents like ChatGPT can also be a valuable tool for program- mers in several practical ways (Timothy 2023). ChatGPT is best used as a companion tool, generat- ing efficient algorithms for specific tasks, answering coding questions, and providing help with ideas. It can also translate code to another language and in certain instances help debug your code. Tyler Lesthaeghe at the University of Dayton Research Institute has been experimenting with the potential FEATURE |AI/ML 30 M A T E R I A L S E V A L U A T I O N • J U L Y 2 0 2 3 2307 ME July dup.indd 30 6/19/23 3:41 PM