
As not many readers of this blog know, my day job is being a product manager in the Israeli high tech industry. And as such, I need to decide how to use modern technology in the products I manage. Today I want to speak about BI vs. AI and what this is all about!
No recent product on the market ever advertised it’s capabilities with „Business Intelligence powered“. If you want to be on top of things, your product has to have „Machine Learning“ (ML) and „Artificial Intelligence“ (AI) as as buzzwords in the description, to gain attention. So simple BI is like the last years best friend we now don’t want to be acquainted with any more. Old fashioned stuff in the fast evolving world of high-tech and big data.
To understand the differences between the two and where Machine Learning (ML) fits in all of this, we need to understand those three terms. Let’s try to explain on very high level what each one means.
Machine Learning (ML)
A machine, more precisely a computer with programmed algorithms, uses automated learning to improve results. Most times it is used to improve accuracy of predictions it makes based on history data. It’s doing so by changing weights of significance for the model the data provides inputs for.
A key part of ML is a validation of predictions. The algorithm needs a feedback about whether the prediction it made was correct. The feedback can be based on manual input, measurement like increased click-through rates or compliance with rules, e.g. in games like Chess and Go: If I won, I made the right decisions on the way.
Predictions can be found everywhere in modern computing. Even within processors that predict upcoming tasks and fulfill them with free cycles to increase overall performance. This is why the value of ML is huge.
Artificial Intelligence (AI)
Intelligence is the capability to make informed decisions based on information, without a precise algorithm. Our brain is performing this in split seconds. Computers working on these tasks use virtual neural networks, that mimic the function of a brain to achieve this. The range of neurons used in real life applications go from as little as two neurons to millions. I once managed a product that used 4 neurons to make informed AI decisions. This is seems very little, but the complexity of the problem it needed to solve was low enough to get extremely good results with 4 neurons.
In many cases, but not all cases, AI works in conjunction with ML and marketing materials often mix up the two.
Business Intelligence (BI)
When working with data, in order to make sense out of it, you need more processing power a human brain can provide. The term „Big Data“ underlines this challenge. But to understand data, you need intelligence. Business Intelligence. This intelligence is provided by the human sitting in front of the BI system, that then allows quick and contextual access to all the data available. The rules how to contextualize the data is not calculated, it is invented and defined by the professional using BI to achieve goals and automate data driven tasks. BI can even be combined with ML without any AI involved.
And yes, we humans are still often better than computers in that. We still understand data better than computers, we just don’t have the power to collect it and analyze it in our brains. This is what we use BI for.
So in many cases, and I am going as far as saying: in most cases, human understanding of data is better than the one of AI. We never start from zero and can create immediate results that make sense and can be used right away. AI can and will take over from there. In the not so distant future. But we as humans and our experience are still needed. Isn’t that a good feeling?
First published on LinkedIn