When it comes to trouble, it is often a result of someone throwing a wrench into the works.
This is certainly the case when it comes to artificial intelligence. However, not in the way we might think.
Up until now, we have engineered machines to be deterministic, which means they are stable across time, reliable, and given a set of inputs, you get the same outputs without variation.
In fact, we spend significant effort to make sure (ensure) there is no variation. This is fundamental to practices such as Lean and Six Sigma along with risk and compliance. All these efforts work to ensure outcomes we want and not the ones we don’t. They make certain that when we do A, we always get B and nothing else.
Yet, here we are, with a stochastic machine, a probabilistic engine we call AI, where the question you ask today will give you a different answer when you ask it tomorrow.
Technically and practically, AI is not reliable, it’s not deterministic. This is not a question of whether AI is accurate or if the answer is correct. It’s about the answer being different every time. There’s always variation in its outputs.
There are many reasons why this is the case, that include the nature of how knowledge models work, the fact that it can learn, and that it can learn how to learn - it can adapt.
However, what is crucial to understand is, AI is not the kind of machine we are used to having in our businesses. We want businesses to be deterministic, predictable, and reliable. And yet here we are,
throwing a stochastic wrench into our deterministic works.
This is why we need to rethink how we govern, manage, and use AI technology.
We need to learn how to be more comfortable with uncertainty. But better than that, we need to learn how to:
improve our probability of success in the presence of uncertainty.