Image Credit: Enterprise Management 360
“The key to artificial intelligence has always been the representation.”
—Jeff Hawkins, Founder of Palm and Handspring
This blog post is authored by Sam Kahn, Data Scientist extraordinaire at Tickr. Prior to Tickr, Sam built machine learning applications for SalesForce.com, Market.Spae, NASA Ames Research Center, and Fullpower Technologies.
Marketers have been hearing a lot about Artificial Intelligence (AI) and Machine Learning (ML) for the past few years, but many are still wondering how they can use these new technologies to capture the hearts and minds of human beings.
This blog post will provide a few concrete tips for utilizing AI and ML to achieve specific marketing objectives. But before we do that, let’s take a step back and look at the bigger picture of the disciplines of AI and ML.
What is Machine Learning?
Artificial Intelligence is basically the science of making things smart. It is the study of agents that perceive the world around them, form plans and make decisions to achieve their goals. Its foundations include mathematics, logic, philosophy, probability, linguistics, neuroscience, and decision theory. Many fields fall under the umbrella of AI, such as computer vision, robotics, natural language processing and machine learning.
Machine Learning is a subfield of artificial intelligence. Quite simply, it uses algorithms to help computers find complex patterns in data, and to learn on their own. This view of machine learning can be traced back to Arthur Samuel’s definition from 1959:
“Machine Learning: Field of study that gives computers the ability to learn without being explicitly programmed.”
Arthur Samuel was one of the pioneers of machine learning. While at IBM, he developed a program that learned how to play checkers better than him. The Samuel Checkers-playing Program was among the world's first successful self-learning programs, and as such a very early demonstration of the fundamental concept of artificial intelligence (AI).
One might wonder why it’s important for machines to learn on their own without being programmed. One reason is that manual programming is slow. Another would be that humans often get things wrong. Traditionally used for making predictions, machine learning can be used to predict a wide range of things, from the outcome of a baseball game to the quality of wines.
Here are a few widely publicized examples of machine learning applications we encounter on a daily basis:
- The heavily hyped, self-driving Google car is the essence of machine learning.
- Online recommendation offers such as those from Amazon and Netflix apply machine learning in our everyday lives.
- Machine learning combined with linguistic rule creation enable us to follow what people are saying on Twitter and Facebook
- Machine learning is commonly used for fraud detection and other forms of online security
Machine learning is capable of analyzing enormous volumes of data – in fact, the more data a model has, the better it will perform. This enables it to do things that would have been cost-prohibitive even a decade ago. In a business setting, ML can be used to predict outcomes such as the likelihood of customer churn, what product features will be most successful, the best path to upsell new services, and more.
Unsupervised Machine Learning
In one type of machine learning, the machine is “unsupervised,” or asked to learn from data that is unlabeled. For example, if unlabeled data about clients is fed into an unsupervised machine learning algorithm, it will naturally group similar clients together into clusters based on their features. These clusters of clients often exhibit similar purchasing habits, upsell potential, and other features. This is one way to segment a business into peer groups.
Tickr’s MetaCloud algorithm is another great example of unsupervised machine learning. We feed it text and it uses complex statistical patterns to extract keywords that are mentioned in the text.
Similarly, our Top Stories algorithm uses unsupervised machine learning to look at news articles to determine how “similar’ they are, and then decides whether they should be clustered together.
Supervised Machine Learning
The other way to approach machine learning happens when the machine is “supervised,” or supplied with training data to learn from. In order to find the correct answer, machines are able to develop algorithms that can start to identify patterns and predict outcomes. But in order for supervised ML to work properly, the data needs to be labeled accurately. This is essential, because the accuracy of the predictions will correspond directly to the accuracy of the labeling.
How can PR practitioners use Machine Learning?
There are a number of ways that machine learning can be used by PR practitioners, as well as other members of the marketing team. Here are three concrete ways:
- Generate Actionable Insights. A machine learning model that uses statistical patterns to extract keywords from text can help to surface insights about customer preferences and behavior. It can also help to illustrate some of the major themes and topics in the traditional and social media activity about a particular industry, company, product, or person. For example, if we use a machine learning model to find news article about a product, it can also tell us what kinds of related sub-topics and keywords are being associated with that product. This can provide valuable insight into the wider context around these stories, and how they relate to the other key phrases and topics in the news. Because machine learning is capable of accommodating huge volumes of data quickly and easily, the insights can be surfaced in real-time, and prescriptive action can be taken immediately. It allows you to contextualize existing coverage, and readjust messaging and outreach strategy on the fly.
- Improve the Signal:Noise Ratio. If you are like most PR practitioners, you spend your day drinking from the firehose when it comes to incoming news and social activity around your brand. Machine Learning can reduce the noise in your news feed by grouping together articles that are similar, and by extracting mentions of specific companies, people, products, or other keywords in those articles. This can provide highly tailored, at-a glance comparisons of trending stories that are most relevant to your brand, along with immediate metrics on reach, engagement and velocity.
- Track customer sentiment. There are many tools out there that help to monitor the quantity of news and social activity, but discerning the quality of that activity can be more difficult. A machine learning model can be used to determine whether a news article is positive or negative, so that brand marketers can keep their finger on the pulse and react appropriately. It can also help provide additional insight about what’s trending, in order to help dynamically shape new content in real time.
—Ginni Rometty, Chairman President and CEO of IBM