Email Jim at jim.thompson@ipulpmedia.com

It would be presumptive of me to say I have a definitive answer to the title today. It is just too early, and many other creative minds must be engaged to answer this question.

However, as a demonstration, I asked ChatGPT this: "What kind of pipe material should be used to transport caustic soda?"

The answer it gave will take all the room I have for the column this week and will thus have to serve as a proxy for the kinds of questions you can ask AI right now:

"Short version: most plants use carbon steel for 50% caustic at ambient temps, and plastics/linings (PVC/CPVC/HDPE/PTFE/PVDF or rubber-lined steel) for dilute solutions or where temps are higher. Avoid aluminum, copper alloys, and galvanized steel.

"Here's a practical guide you can use right away:

"Good choices (pick by concentration & temperature)

"Carbon steel (welded, not galvanized)

"Great for ~20-73% NaOH at ambient to ~60 °C (140 °F).

"Add corrosion allowance; watch for caustic stress-corrosion cracking (SCC) if hot (>60 °C) or under high stress.

"CPVC / PVC / PE (HDPE) / PP

"Excellent for dilute to ~50% at moderate temps.

"Rough limits: PVC ≤60 °C (140 °F), CPVC ≤90-95 °C (≈200 °F), HDPE ≤50-60 °C, PP ≤80-90 °C (vendor ratings vary).

"PTFE / PVDF

"Broad chemical resistance and higher temp capability; use for hot caustic or where you want a long-life nonmetal solution (lined steel or all-plastic).

"Rubber-lined (e.g., EPDM) or FRP (vinyl-ester resin) pipe

"Common for dilute caustic and moderate temps, especially large diameters or corrosive environments.

"Nickel alloys (e.g., Alloy 200/201, Monel 400)

"Best for hot, strong caustic where steels and stainless suffer SCC; usually a costly last resort or for critical sections.

"Use with caution / generally avoid

"Austenitic stainless (304/316): OK for cold, dilute service; prone to caustic SCC at higher concentration/temperature--often a poor choice for strong/hot caustic.

"Aluminum, copper, brass/bronze, zinc/galvanized: Do not use--they're attacked/dissolved by NaOH.

"Threaded connections in metal caustic service: prefer welded (steel) or heat-fused (thermoplastics) to minimize leak points.

"Seals, gaskets, and valves

"Gaskets/linings: PTFE is the safest default; EPDM is typically compatible.

"Avoid natural rubber, FKM/Viton in strong hot caustic (tends to degrade).

"Valves: Lined plug/ball valves (PTFE/PFA), alloy trim where needed; keep stems/packing compatible (PTFE/EPDM).

"Heat tracing & operations

"50% caustic crystallizes around ~12 °C (54 °F) → insulate and heat trace in cool climates.

"Don't use copper heat-trace; use steam, stainless, or electric trace compatible with caustic.

"Provide drains/flushes and allow for thermal expansion in plastics.

"If you can share concentration and temperature (and if it's indoors/outdoors), I'll pin it down to a specific material spec (e.g., "Schedule 40 carbon steel with X mm corrosion allowance" vs "SCH80 CPVC with solvent-welded joints," etc.)."

While this appears to be magical, there are a couple of severe problems. First, sources are not cited, so we don't know if this set of answers has the proper vetting. Second, the sources have no way of earning an income from their work, there are not even advertisements. Over time, this will degrade the material available, for who wants to do such work for free? This will be settled in the courts in both cases. The first, in the case of catastrophic failures and the second in cases of copyright and intellectual property.

In the meantime, try questions of your own, but be careful about applying the answers to real life problems.

Be safe and we will talk next week.

For a deeper dive, go here.

AI in Capital Projects: A Study Guide

I. Core Concepts & Themes

This document explores the emerging role of Artificial Intelligence (AI), specifically large language models like ChatGPT, in the context of capital projects. The central theme revolves around the potential benefits and significant challenges associated with using AI for practical engineering and material selection tasks.

Key Learning Objectives:

• Understand the current capabilities of AI in providing technical information.
• Identify the critical limitations and risks of relying on AI-generated content in capital projects.
• Recognize the specific concerns regarding source citation, intellectual property, and financial models for content creators.
• Grasp the practical recommendations for selecting pipe materials, seals, gaskets, and valves for caustic soda transport, as generated by AI.

II. Detailed Content Review

A. AI's Role and Capabilities

Initial Assessment: The author, Jim Thompson, states it's "too early" to definitively answer where AI fits in capital projects, emphasizing the need for broader engagement.

Demonstration: ChatGPT was asked, "What kind of pipe material should be used to transport caustic soda?"

AI's Output - Material Selection for Caustic Soda: Short Version: Carbon steel for 50% caustic at ambient temperatures; plastics/linings (PVC/CPVC/HDPE/PTFE/PVDF or rubber-lined steel) for dilute solutions or higher temperatures. Avoid aluminum, copper alloys, and galvanized steel.

[truncated]

Specific Recommendations: AI notes it could provide more specific material specs if concentration, temperature, and indoor/outdoor conditions are provided.

B. Problems and Limitations of AI

Lack of Source Citation: AI does not cite its sources, making proper vetting impossible. This raises concerns about the accuracy and reliability of the information.

Financial Model for Content Creation: Sources (human experts/creators) have no way of earning income from their work when AI aggregates and presents it.

Absence of advertisements or direct compensation will "degrade the material available" over time as creators are disincentivized.

Legal Implications: Catastrophic Failures: Lack of vetting and potential for incorrect information could lead to severe engineering failures, resulting in legal battles.

Copyright and Intellectual Property: The use of existing content without attribution or compensation raises significant copyright and intellectual property issues, which are expected to be settled in courts.

Cautionary Note: The author advises trying AI for questions but warns against "applying the answers to real life problems" due to the aforementioned issues.

III. Glossary of Key Terms

• Artificial Intelligence (AI): The simulation of human intelligence processes by machines, especially computer systems. These processes include learning, reasoning, and self-correction.
• Capital Projects: Long-term investments in physical assets, such as infrastructure, buildings, or industrial plants, often involving significant capital expenditure.
• Caustic Soda (NaOH): Also known as sodium hydroxide, a highly corrosive chemical widely used in various industries.
• ChatGPT: A large language model developed by OpenAI, capable of generating human-like text responses to a wide range of prompts.
• Corrosion Allowance: An additional thickness of material added to a pipe or vessel beyond that required for design pressure, to account for anticipated material loss due to corrosion over its design life.
• Caustic Stress-Corrosion Cracking (SCC): A form of environmental-assisted cracking that occurs in certain metals (e.g., carbon steel, austenitic stainless steel) exposed to specific corrosive environments (like hot, concentrated caustic soda) while under tensile stress.
• HDPE (High-Density Polyethylene): A thermoplastic polymer made from the monomer ethylene, known for its high strength-to-density ratio, used in piping.
• FRP (Fiberglass Reinforced Plastic): A composite material made of a plastic matrix reinforced with fine fibers of glass. Used for its strength and corrosion resistance.
• Monomer: A molecule that forms a polymer chain; many monomers linked together form a polymer.
• NaOH: The chemical formula for sodium hydroxide, commonly known as caustic soda.
• PFA (Perfluoroalkoxy): A fluoropolymer with properties similar to PTFE but with better melt processability.
• PP (Polypropylene): A thermoplastic polymer used in a wide variety of applications, including piping, known for good chemical resistance.
• PTFE (Polytetrafluoroethylene): A synthetic fluoropolymer of tetrafluoroethylene, known for its non-stick properties and high chemical resistance (e.g., Teflon).
• PVDF (Polyvinylidene Fluoride): A high-performance fluoroplastic known for its chemical resistance, thermal stability, and mechanical strength.
• PVC (Polyvinyl Chloride): A widely used plastic polymer, known for its durability and relatively low cost, common in piping.
• CPVC (Chlorinated Polyvinyl Chloride): A thermoplastic produced by chlorination of polyvinyl chloride resin, making it more flexible and capable of withstanding higher temperatures than standard PVC.
• Thermosetting Plastic (Thermoset): A plastic that, once cured by heat or other means, is permanently rigid and cannot be softened and reshaped by reheating.
• Thermoplastic: A plastic polymer that becomes pliable or moldable above a specific temperature and solidifies upon cooling. It can be reheated and reshaped repeatedly.
• Vetting: The process of carefully investigating someone or something before making a decision or taking action, to ensure reliability and suitability.

Quiz: AI in Capital Projects

Instructions: Answer each question in 2-3 sentences.

1. According to Jim Thompson, what is the current state of definitively answering where AI fits into capital projects?
2. What specific question did Jim Thompson ask ChatGPT to demonstrate AI's capabilities for material selection?
3. Name two pipe materials that ChatGPT recommends for transporting 50% caustic soda at ambient temperatures.
4. Why does ChatGPT suggest avoiding aluminum, copper, and galvanized steel for caustic soda service?
5. What significant problem does the author identify regarding the sources of AI-generated information?
6. Explain how the current financial model (or lack thereof) for AI-generated content might degrade the quality of available information over time.
7. What is caustic stress-corrosion cracking (SCC), and in which materials is it a concern for caustic service?
8. Which two materials are generally recommended by AI for gaskets and linings in caustic service?
9. Why is heat tracing important for 50% caustic soda in cool climates, according to the AI's response?
10. What are the two primary legal issues the author anticipates will arise concerning AI-generated content?

Quiz Answer Key

1. According to Jim Thompson, it is "too early" to provide a definitive answer on where AI fits into capital projects. He emphasizes that many other creative minds need to be engaged to properly address this complex question.
2. Jim Thompson asked ChatGPT, "What kind of pipe material should be used to transport caustic soda?" This question served as a practical demonstration of the types of inquiries AI can currently answer.
3. ChatGPT recommends carbon steel for ~50% caustic at ambient temperatures. Additionally, plastics like PVC, CPVC, HDPE, or PP are suggested for dilute to ~50% solutions at moderate temperatures.
4. ChatGPT suggests avoiding aluminum, copper, brass/bronze, zinc, and galvanized steel because these materials are actively attacked or dissolved by sodium hydroxide (caustic soda). Their chemical incompatibility makes them unsuitable for this service.
5. A significant problem identified by the author is the lack of source citation in AI-generated information. Without knowing the sources, it's impossible to properly vet the answers for accuracy and reliability.
6. The current financial model, where content creators receive no income or advertisements from AI utilizing their work, will disincentivize experts from creating and sharing high-quality information. Over time, this lack of compensation is predicted to degrade the overall quality and availability of source material.
7. Caustic stress-corrosion cracking (SCC) is a form of material degradation that occurs when certain metals are exposed to specific corrosive environments (like hot, concentrated caustic) while under tensile stress. It is a particular concern for carbon steel if hot (>60 °C) or under high stress, and for austenitic stainless steels (304/316) at higher concentration/temperature.
8. For gaskets and linings in caustic service, PTFE is presented as the safest default choice. EPDM is also typically considered compatible and is recommended by the AI.
9. Heat tracing is important for 50% caustic soda in cool climates because it crystallizes around 12 °C (54 °F). Insulation and heat tracing prevent solidification, which could block pipes and disrupt operations.
10. The two primary legal issues the author anticipates are, first, cases involving catastrophic failures resulting from incorrect AI-generated advice, and second, cases related to copyright and intellectual property infringement due to AI's use of existing content without proper attribution or compensation.

Essay Format Questions

1. Discuss the dual nature of AI's current capabilities in engineering tasks, using the caustic soda pipe material selection as an example. Highlight both the apparent "magical" utility and the critical underlying problems identified by Jim Thompson.
2. Analyze the long-term implications of AI's lack of a financial model for content creators. How might this impact the quality, availability, and reliability of the technical information that AI models draw upon in the future?
3. Evaluate the legal ramifications discussed in the article, focusing on both catastrophic failures and intellectual property. Explain why these issues are significant challenges to the widespread adoption of AI in critical capital projects.
4. Beyond the specific material recommendations, what general principles or best practices for material selection and design can be inferred from the AI's response regarding caustic soda transport? Consider factors like temperature, concentration, and connection methods.
5. Imagine you are an engineer tasked with a capital project involving caustic soda. Based on Jim Thompson's article, outline a responsible approach to using AI for preliminary material selection, including steps you would take to mitigate the identified risks before applying any AI-generated recommendations to real-life problems.

________

Other interesting stories: