In a world of AI, how should RDS adopt?

Reference designations have had a long history and the requirements to them are reflected in technological advancement.

Initially reference designations were used primarily on electrical diagrams and circuit boards to associate attributes to the components. The system had human readability in focus as it was a manual task to look up the codes, interpret them and trace them through different documents.

With only one engineering field in mind and simple systems, single letter codes were used to identify the type of component, and numbers to make each instance unambiguous.

As complexity grew, new engineering fields had to coordinate, and that meant the simple one letter system could not scale. Electrical engineering had to coordinate with mechanical-, process- and software-engineering as well. Components had to be organized according to functionality and location.

One letter classification codes became three letter classification codes, products became functions, components became systems.

This helped advance the digital age where assets are traced not only for maintenance and operations, but concept, design, and retirement as well. RDS helped push the digital thread in organizations.

Now we stand in a new digital age where AI is advancing quickly, the technological advancement is discussed daily and the forecast is updated just as often. The capability of AI seems to be ever growing. We must ask ourselves as experts: should the RDS principles be updated to reflect the new technology the same way one letters where updated to three letters?

In the new AI world the human interface in communication has been vastly reduced. There is little need for the human brain to understand how or why the answer to a question is correct, as long as it is correct.

RDS has always been designed to be verified, constructable and usable for human beings, to communicate information through a uniform syntax to any IT-system.

AI changes that. The rigorous syntax which has been important for IT systems for such a long time, is now just a “nice option”. AI will eat your data for breakfast, no matter how messy it is, and give you a valid answer – as long as the amount of data is big enough and the context is clear. Why would anyone bother with aspects, classification codes, or syntax rules?

This question is not only relevant for the RDS domain. The answer becomes more clear when you realize it applies to many other industries as well. The software development world will be the first to test whether or not a human layer is necessary. All development languages are designed to be human readable and verifiable, but if the AI is writing all the code anyway, why did it need to be human readable in the first place?

We don’t know the answer to this for sure yet, but it seems obvious no organization would run critical software without having the human verification aspect layer on top of it. An AI does not care if you fire it, you cannot sue it if breaches contract or does not do as instructed – it cannot be held accountable, only humans can.

With this in mind it is much more probable that AI will be used to transform, classify and review RDS structures and translate messy data into RDS, than it is AI will replace the RDS structures all together. AI will use RDS 81346 as the language to structure designs and communicate their thoughts and intentions to humans. Human experts will verify and validate their answers and make sure it does not conflict with the organization strategy or needs from stakeholders.

RDS will play a major role in constraining an AI and making sure it does not hallucinate. RDS defines the boundaries for its outputs and is there to explain the Ais decisions and reject its proposals.

If we assume Reference designations and the principles that governs them in IEC 81346 will stay, then how is the system updated to fit the future? We believe the relevance of being a digital thread becomes even more important in the future. This means not only working with a single tree structure and a single aspect, but multiple designations, relations and datapoints for an object which can be tied together in a graph. Definitions and semantics behind every class code, common attributes and dependencies will be data-gold for an AI, to provide a more precise and probable answer.