Demystifying AI to enable strategic decision-makers
The idea of AI is ubiquitous in the evolving decision-making landscape today, and as we’ve written about before, is quantifiably widening the gap between high performers and those who are lagging behind digitally.
Researchers have looked at how AI in business has changed over the last ten years from a new digital trend to a vital technology now helping high performers make data-driven decisions within organizations (Kitsios and Kamariotou, 2021). Yet there continues to be a knowledge gap in how to strategically apply AI for new business creation. This gap is likely persistent because those with the business expertise to see effective patterns of application are still not well-versed in the underlying mechanics of artificial intelligence and machine learning.
In this article, we will attempt to take some of the technical aspects of artificial intelligence and help decision-makers better understand how AI can be integrated into their network and what possibilities exist within their organization.
Machine Learning and Artificial Intelligence in Business
First off, we can discuss the difference between machine learning and artificial intelligence (AI). Often these terms are used interchangeably, but they are better understood as a kind of Russian nesting doll. AI encompasses several techniques to mimic intelligence, including machine learning.
The “father of AI”, John McCarthy, describes it as the science and engineering involved in creating intelligent machines, particularly intelligent computer programs (Kumar & Chandrakala, 2016). What do we mean by intelligence? This typically means that the programs themselves are capable of replicating human behaviors by studying human thought patterns and problem-solving methods.
Machine learning, then, is a type of AI that allows computers to develop their own predictions and recommendations based on large amounts of raw data. If you’ve ever heard of the now-famous experiment where a computer beat a master at the Chinese game of Go, this feat was performed via machine learning. Whereas artificial intelligence would have simply replicated the thought patterns of the human across the table, machine learning allowed for the computer to come up with a unique solution to the problem that allowed it to beat its extremely adept Go opponent.
We can think of these differences in business through separate use cases. Consider your last interaction with a chatbot on a website. The chatbot likely attempted to speak to you and substitute wholly for a human on the other side of the screen. You can think of the overarching backbone of a chatbot as AI, which has been trained via machine learning and other AI methods.
Machine learning is demonstrated well by a use case involving training computers to synthesize thousands of radiology reports and notes to be able to detect the earliest signs of breast cancer.
When an AI program trained through machine learning is used alongside a human assessor, studies have found that breast cancer can be detected at a 20% higher rate.
As is evident in the example above, machine learning and AI are often used synonymously because they work in concert with one another to aid business and healthcare decisions. AI simply means that there is an artificial intelligence, or computer-generated decision-making matrix, operating in an infrastructure. This AI may or may not have been trained using machine learning.
What about Neural Networks in Business?
Another buzzword you may have heard in your interactions with AI is “neural network.” Just as machine learning is nested under the AI umbrella, a neural network is nested a couple layers within machine learning. The next level under machine learning is deep learning, a subset of machine learning wherein there is much less human input involved. Deep learning uses neural networks to perform this type of machine learning.
In our breast cancer example above, humans were involved in labeling the data, correcting erroneous outputs as the machine learning of the AI took place, and were the final assessors and correctors of the AI’s conclusions. This differs from deep learning, wherein the computer would not require the same amount of human hand-holding to learn how to make decisions and draw conclusions.
Neural networks create layers of inputs that are processed into an output. At the very top of the pyramid is the data that a machine learning algorithm studies, and then beneath are the neural networks populated with nodes that are meant to mimic neurons. Think of each neural network as a complex set of algorithms that examine data, assign weight to that data, and then use the combination of that data to populate a node’s value. This combination of values then travels down to the next neural network layer to make more decisions.
The neural networks allow deep learning to take place by training a computer on what raw data represents in the real world by weighing various data points against one another. For example, a neural network would use an algorithm to weigh financial transactions across a network either by their value, location, or any other data type assigned a value and weight in that algorithm. Then, it would pass that information onto the next part of the deep learning network to evaluate that data against the next set of algorithms needed to make a decision. The combination of this data processed through the entirety of these neural networks can then alert a company to possible fraud. Deep learning simply means that there are stacks of these neural networks used to make a decision rather than two or three levels involved in simpler machine learning.
Conclusion
If you’re not sure of what kind of magic happens to make a computer create new algorithms in its own neural network or optimize its decision-making ability, you are not alone. Even those who are artificial intelligence experts will tell you that some deep learning is still a mystery to those who create the programs.
What you need to understand to get the most out of artificial intelligence in business is that there are important decisions being made by human programmers who create the algorithms and parse the data that trains a computer on machine learning.
The benefit of working with an AI consultant is that these algorithms are refined and created based directly on the feedback your team gives on how data is typically weighted in existing algorithms, or how it should be weighted in new areas of technological adoption in your business AI.
If you’re interested in understanding how this technology can help grow your business and make you and your team more efficient, more data-backed, and more productive, reach out today!