AI and all that Jazz

Paraic O’Toole, CEO Automsoft, on the value of AI and all that jazz.

AI and predictive analytics is a field that is attracting an increasing amount of attention. Like every new field that emerges, there is a lot of confusion, conflicting definitions, claims of transformational industry change and marketing fluff wrapped around a technology development that can indeed transform our environment. When you add in Deep Learning into the mix the picture becomes both more exciting and more confusing. 

I recently spoke to a CEO of a campus incubation company which is developing great technology embedding deep learning into IoT edge networks. He was frustrated at the amount of time he had to spend educating VCs he was pitching to. He was also frustrated at the lack of knowledge of the space on the part of supposed subject matter experts hired by the VCs to evaluate the technology. Like everyone else who is honest, I am learning but what I am seeing is really exciting.

There are differences with Artificial Intelligence and Machine Learning and then there is Deep Learning. There are anomaly detection algorithms and self-learning models. First, the starting point is that we have to think differently, this is not your run of the mill software. Rather than plug and play, this is a product that evolves after installation. Each system needs a training data set to refine and train the model. Some customers don’t appreciate the idea that the system is learning to be effective six months after install, and that that time period is proportionate to the uniqueness of the environment.

However, rather than go into the detail of ‘what is AI’ which, though attractive is of limited value, it is more relevant to ask what benefits are we looking for from AI. Predicting failures is a clear business benefit, but is a spectrum. A prediction can be made based on straight line extrapolation of factors like vibration but that isn’t really AI. Applying a ML algorithm will enable a more variable operating profile to be assessed and failure predicted and the more complex the operating environment, the more complex the AI required to predict a failure. However, this is really a collaborative process between AI people and the client and the most important factor is not the technology but the business impact. 

It seems strange to be reiterating one of the most basic tenets of buying technology but when something like AI comes along, common sense business principles can get discarded in favour of the seductiveness of the technology. The most important question, then, is not how does it work, or even how much does it cost, but what does it deliver to the business? It really is that simple.