Artificial Intelligence (AI) has revolutionized many industries, offering unprecedented efficiency, scalability, and innovation. In the field of nondestructive testing (NDT), tools like ChatGPT have shown great promise in generating documentation, simplifying complex concepts, and even assisting with procedural development. However, as an eddy current testing (ECT) expert, I feel compelled to highlight the potential dangers of relying too heavily on AI to write NDT procedures—especially in a field where precision, safety, and tacit experience are paramount.
What ChatGPT Does Well
First, let’s acknowledge where AI excels. ChatGPT and similar tools are fantastic at generating clear, structured, and grammatically correct content. They can quickly compile information from vast datasets, making them useful for drafting general guidelines, summarizing standards, or creating training materials. For example, if you need a basic outline of an eddy current testing procedure for a common application like tubing inspection, ChatGPT can provide a solid starting point.
AI is also excellent at translating technical jargon into more accessible language, which can be helpful for training new technicians or communicating with stakeholders who may not have a deep understanding of NDT. Additionally, it can assist in identifying relevant codes and standards (e.g., ASME, ISO, or ASTM) that apply to a specific testing scenario.
Where Caution is Critical
Despite these strengths, there are significant risks in using AI to write NDT procedures without human oversight. Here’s why:
Lack of Tacit Knowledge:
NDT, and particularly eddy current testing, relies heavily on tacit knowledge—the kind of expertise that comes from years of hands-on experience. This includes understanding how subtle variations in material properties, surface conditions, or probe positioning can affect test results. AI lacks this tactile, experiential understanding. It can’t “feel” the difference between a good signal and a false indication, nor can it intuitively adjust for real-world variables that aren’t explicitly documented.
For example, an AI might generate a procedure that specifies a particular frequency for detecting cracks in a specific alloy. However, it may not account for the fact that the potential damage mechanisms are expected to occur at material edges. Only an experienced technician would recognize the need to used a shielded coil and perhaps special probe handling (like a plastic probe fixture) based on these factors.
Overreliance on Generic Data:
AI models like ChatGPT are trained on publicly available data, which may not always reflect the latest advancements, proprietary techniques, or nuanced best practices in NDT. This can lead to procedures that are technically correct but suboptimal or even unsafe in practice. For instance, an AI-generated procedure might not account for the unique challenges of inspecting a complex geometry or a material with anisotropic properties.
3. Inability to Handle Edge Cases:
NDT often involves dealing with edge cases—unexpected flaws, unusual material behaviors, or challenging environmental conditions. AI struggles to anticipate these scenarios because it lacks the ability to reason creatively or draw on real-world experience. A human inspector, on the other hand, can adapt on the fly, using their tacit knowledge to troubleshoot and refine the procedure as needed.
4. Ethical and Liability Concerns:
If an AI-generated procedure leads to a missed defect or a catastrophic failure, who is responsible? The technician? The company? The AI developer? This gray area poses significant ethical and legal challenges. NDT is a critical safety function in industries like aerospace, energy, and manufacturing, and the stakes are simply too high to rely on unverified AI-generated content.
The Role of Tacit Experience in NDT
Tacit experience is the backbone of effective NDT. It’s what allows an inspector to interpret ambiguous signals, recognize patterns, and make judgment calls that go beyond the written procedure. For example, in eddy current testing, an experienced technician might notice a slight phase shift in the signal that suggests a subsurface defect, even if the amplitude is within acceptable limits. This kind of insight is impossible to codify fully in a written procedure, let alone replicate with AI.
Best Practices for Using AI in NDT
While AI can be a valuable tool, it should never replace human expertise in NDT. Here are some best practices for integrating AI into your workflow:
Use AI as a Starting Point, Not a Final Product:
Treat AI-generated procedures as drafts that require thorough review and validation by experienced NDT professionals. This ensures that tacit knowledge and practical considerations are incorporated.
Combine AI with Human Oversight:
Pair AI tools with seasoned inspectors who can identify gaps, refine procedures, and adapt them to real-world conditions. This hybrid approach leverages the strengths of both AI and human expertise.
3. Validate Procedures Through Testing:
Before implementing an AI-generated procedure, conduct rigorous testing to verify its accuracy and reliability. This is especially important for critical applications where failure is not an option.
4. Invest in Training and Knowledge Transfer:
Use AI to support training and knowledge transfer, but ensure that new technicians gain hands-on experience and learn from seasoned professionals. Tacit knowledge is best passed down through mentorship and practice.
Conclusion
AI tools like ChatGPT have the potential to enhance the field of nondestructive testing, but they must be used with caution. Writing NDT procedures is not just about following a set of rules—it’s about applying deep, often tacit, knowledge to ensure safety, accuracy, and reliability. As an eddy current expert, I urge the NDT community to embrace AI as a complementary tool, not a replacement for human expertise. After all, when it comes to ensuring the integrity of critical infrastructure, there’s no substitute for the judgment and experience of a skilled inspector.
Let’s use AI wisely, but never forget the value of the human touch.
For more information on NDT and Eddy Current Testing, visit eddycurrent.com, your one-stop-shop for eddy current testing.
Comments