3 Subsets of Artificial Intelligence

3 Subsets of AI

Gartner reported that companies deploying Artificial Intelligence (AI) grew from 4% to 14% between 2018 and 2019. As its applications and capabilities spread, many companies are looking for more ways to use AI in their businesses, but are struggling to understand how and why. News about AlphaGo beating a human in playing Go, or the latest robot from Boston Dynamics, looks interesting and innovative, but are these AI applications useful for business? 

Unfortunately, many business leaders and analysts struggle to understand AI. They see the technology with a narrow scope and envision Chatbots and Automation. It’s reasonable for these business leaders to think of such examples due to the fact that Chatbots are a familiar concept to most, and fairly easy to implement, while Automation is trending and therefore provides good PR. The true power of AI, however, is very simple. 

What is AI?

Let’s consider a very simple example: If you went to the bathroom today, you might see the machine that dispenses paper towels. That’s AI. AI is anything artificial that mimics intelligent behavior. If you ask someone in the bathroom to hand you a paper towel, you will expect that person to give you a paper towel, because humans are intelligent beings. If you asked a lobster to give you a paper towel, you wouldn’t expect a response, certainly not actually receiving your paper towel. 

When you “ask” that paper towel dispenser to give you a paper towel by putting your hands in a particular place, it mimics intelligent behavior. It responds to the request by dispensing a paper towel. It’s a semblance of intelligence in that very narrow experience. That’s what AI really is: A very simple technology that mimics some form of intelligent behavior. 

With this definition of what AI really is in mind, let’s discuss AI in terms of business. 

Why use AI in business?

Artificial Intelligence can do many things, but in business, it becomes powerful when applied to data analytics. AI is extremely powerful and fast at searching for patterns in data, finding relationships, and testing the assumptions when new datasets are given. For the most part, AI doesn’t need any human interaction. With AI, any business can analyze large complex datasets, find patterns and relationships, and make predictions to improve their business. 

AI is a broad field, and as that field has progressed, several subsets have emerged that describe applications of AI. Understanding the utilization of Machine Learning, Neural Networks, and Deep Learning individually will help you understand the big picture.

AI and its 3 subsets

Machine Learning

The first subset of AI is Machine Learning. Machine Learning is a method that is used to achieve AI by giving a machine a set of data and asking it to learn to predict a real-life outcome. Within Machine Learning, you can have Supervised or Unsupervised Machine Learning, based on the availability of a target variable, or the output.

Supervised Machine Learning

When we have the output values, we can teach the machine to learn the relationship, or find the pattern, between the input and the output values, thus allowing it to find the best way to build a model that will predict an output when it’s given a new set of input values. Let’s go back to our paper towel dispenser and use Machine Learning to make it smarter. With Machine Learning, you can have the paper towel dispenser predict if the hand waving at it belongs to an adult or a child, and adjust the length of paper towel accordingly. Adults might need longer paper towels for bigger hands; accordingly, shorter paper towels for child-sized hands. A dispenser using Machine Learning would have the data for the ideal length of paper towel for different sizes of hands. From there, it can use that data to learn what length of paper towel it should dispense, and adjust the length accordingly.

Another usage example of Machine Learning could be a machine that predicts the probability of a person developing diabetes. If you’re looking at different key parameters such as height, weight, BMI, age, blood sugar levels, and genetic markers, you can see that there’s no way to write a simple code. Every combination of those parameters is a massively huge number, bordering on infinity. With Machine Learning, the machine will use the data available to learn the rules and predict if an individual has a high risk of developing diabetes. So when data for a new case is input, the machine will use what it built internally to make a prediction regarding the probability of that person developing diabetes.

In technical terms, Supervised Machine Learning works well with classification and regression, and some well-known algorithms include logistic regression, decision trees, and support vector machines. 

Unsupervised Machine Learning

Unsupervised Machine Learning is for situations where there isn’t a specific set of output labels, with a goal of learning from the structure of the data. One common usage of unsupervised learning is clustering – grouping data points that are similar to each other without labeling them. For example, you could utilize Unsupervised Machine Learning to segment your customers more effectively by allowing the machine to find patterns that are complex to identify.

Neural Networks

Neural Networks can be used in both Supervised and Unsupervised Machine Learning. While there has been a lot of focus on, and success with image data sets and detecting objects, Neural Networks can be used with anything. Simply put, Neural Networks are simply a series of additions and multiplications. It’s an advanced way of doing a simple function.

Neural Networks are designed to learn from data with clearly defined stages of observation. They are made up of stacked layers of simple computational nodes that work in unison to find patterns within the data. Neural Networks are trained by being fed a large set of labeled data. Let’s say you want to build a machine that can recognize the difference between elephants and rhinos. You would give the Neural Network animal pictures of elephants and rhinos paired with their corresponding animal names, such as this is an elephant’s ear, this is a rhino’s horn, etc. The Neural Network then goes to work to solve a complex mathematical puzzle. It takes the data of the picture-name pair and learns a formula that turns the image into the category. After it solves or learns that puzzle, it can now reuse the formula to correctly categorize any new animal photo as either an elephant or a rhino.

A single formula to describe the picture-to-name process would be overly broad, and result in a low-accuracy model, giving you false negatives and positives. This is where the Neural Network’s layers come into play. Layers break the process into different steps that allow the network to find a series of formulas that describe each stage of the process. The first layer might take in all the pixels, and use a formula to pick out which ones are most relevant to elephants or rhinos. The second layer uses larger patterns from grouping a bunch of pixels, and figures out whether the image has floppy ears or horns. Each subsequent layer would identify increasingly complex features of the animal until the final layer decides the specific animal based on the cumulative calculations. By breaking down the process step-by-step, the Neural Network can build more sophisticated models, which leads to more accurate predictions.

Deep Learning

Neural Networks are at the core of what makes Deep Learning so powerful. Deep Learning is another application of Machine Learning that is particularly useful when working with unstructured and unlabeled data. It adds layers of complexity – going “deep” – to what you’re asking your machine to learn, with the ability to learn from its mistakes and deliver probability with a degree of accuracy. As a result, Deep Learning is commonly used in analyzing images, videos, and speech recognition.

As we can see, different techniques may be necessary to solve business problems correctly. The more we understand AI, the closer we will become to utilizing AI to impact business outcomes. If you aren’t sure whether AI is for you, please read this article that was written in partnership with VentureBeat. If you are ready to discuss how to implement AI in your business, contact us today!