A Seth Godin blog and Peter Drucker daily digest one day last week brought me back to my IEC 62443’s Future … Encyclopedia Brittanica and AI article, originally published in January 2023. The article raised the ire of many on the 62443 committee, primarily around the conservative consensus claim. In twenty years of writing weekly articles, I’ve found that often the main point gets set aside, and this was the case here.

My point was the length, complexity and effort required to get value from the IEC 62443 standards reduces the use of these standards. And it will lead to a decreased use and impact of the standards in the future given the trend line of how information is gathered and used. One way to turn this around was something I referred to, a bit tongue in cheek, as Chat62443.

The value of the 62443 documents, their impact on creating a reliable and secure IACS environment in all sectors, globally, would be greatly improved if a Chat62443 could answer questions based on the 62443 conservative consensus body of knowledge.

Seth Godin is playing around with a ChatSeth. From Seth Godin’s March 29, 2023 post:

An experimental AI chat bot that has been trained on all 5,000,000 words of this blog. You can find it here. Yes, you can trick it, but you can also ask it questions about anything I’ve blogged and it may do a good job of answering.

The challenge is there may not be enough text in the 62443 documents for a LLM. I have no idea what the word count is. There were some suggestions, that other ICS standard and guideline documents could be included. There may be some conflicts, although most of the conventional wisdom in these various documents is consistent and LLMs are always dealing with conflicting data. It would no longer be Chat62443 then, more of a ChatOTSEC. 

If the 62443 writers had focused on writing more FAQ’s, guides based on subsections, and other supporting content there would be more content to train the model. 

I can guess on two objections to Chat62443. The first is best stated in a comment to my early article by Khalid Ansari:

Chat62443 is a tongue-in-cheek idea that trivializes the value of standards. One really must understand the standard wholly to benefit fully. 

This aligns with my earlier point that requiring one to read and understand multiple standards is a high barrier to entry and risks decreased relevance. 

And second, a Chat62443 will not always be right, in the sense it may answer something different than what was suggested or required in the 62443 standards. This is understandably troubling given the years of labor involved in writing and editing these standards. 

Can the 62443 team live with their solution not being near perfect and being much more widely used? Or would they prefer to keep in near perfect, in the authoring committee’s mind, and much less used due to the need to access and understand multiple standards to get an answer?

This is where the Drucker quote comes in. From Peter Drucker’s Landmarks of Tomorrow:

In some cases the best way to strengthen the system may be to weaken a part — to make it less precise or less efficient. For what matters in any system is the performance of the whole;

Do we measure the performance of 62443 on the perfection of the text or on the impact on asset owners securing their ICS?

Again I’d make the case the Chat62443 could enliven the standards, and increase the usage of all the hard work that went into writing them. It may even improve that body of knowledge. Wrong or imprecise answers, to relevant questions, created by Chat62443 could trigger articles on the right answer that could be fed back into the model. It may even identify areas where the language could be clarified and inconsistencies exist. 

Why not give it a try, experiment, a la Seth Godin?

One last thing … Chat62443 could also help with the business model. Although this is a lot more complicated given the decision long ago to tie ISA99 to IEC. Not a wrong decision, but it likely requires more approvals and bravery.