The AI RMF is here
It’s rare to expect anything exciting to come from government agencies. But, on occasion, you might get pleasantly surprised. On January 26, the National Institute of Standards and Technology (NIST) released the Artificial Intelligence Risk Management Framework (AI RMF 1.0).
This project is a culmination of 18 months of drafting, commenting and refining the framework which should become the basis for deploying any technology under the label of AI.
Why do I think that this is an important event?
The AI RMF publication addresses issues like risk and trustworthiness. It is defining what it means for artificial intelligence to be valid and reliable. What does this really mean? It’s about making AI safe, transparent, and easy to explain.
Currently the framework — and any compliance with it — is purely voluntary. It will take another few years for the framework to mature. The technology around AI is changing at an increasing speed and you can expect revisions coming at a relatively fast pace.
My estimate is that, at the beginning, it will be mainly used by marketing departments. Why? If you’re a younger reader, you won’t remember what happened in the past — any new product introduced to the market has to start with 'i'. This was to reference the word ‘internet,’ which is why we have the iPhone or iPad. These days if your product doesn't have 'AI' in the description, it is considered a 'blast from the past'.
How can you tell that this product has better AI than the other? Your product will be more safe, more transparent, more explainable than your competition.
It will become a race which ChatGPT will happily help you compete in. Marketing people will come to all the data engineers and endlessly ask them if they can declare their product the most valid, transparent, explainable and reliable in the market. Yes, the immediate result will be the complete disdain from tech people towards marketing (and you thought it couldn't get worse).
In time, all the marketing claims will have to be supported by numbers; eventually, by verifiable numbers. These numbers will have to be valid, reliable, explainable and transparent. Any bias will be spotted and, without mercy, that criticism will be thrown back at the competition. The marketing and data departments will have to learn how to work together in a way that is not common today.
'AI' technology is slowly maturing. We are learning how to build it. We are trying to understand where it can help us and where it is best not to use it. We are making efforts to make it useful for us. It is the same recurrent pattern with everything else we've done in the past. Exciting times!