The Digital Renaissance And Machine Learning: Embedding Easy-To-Consume Business Capabilities To Solve Common Challenges

Ginger Shimp and Jeff Janiszewski

The following is the fifth in a series of conversations about digital innovation and the intelligent technologies powering the Intelligent Enterprise, with Jeff Janiszewski and Ginger Shimp from SAP North America Marketing. In this blog they discuss how machine learning can be used to find undiscovered data.

Ginger Shimp: So, are you looking forward to NaNoGenMo?

Jeff Janiszewski: I don’t know what that is.

GS: It’s short for “National Novel Generation Month,” and it happens every November. It was modeled after “National Novel Writing Month,” in which participants have 30 days to write a 50,000-word novel. However, NaNoGenMo challenges participants to create a computer program that will write a 50,000-word novel in one month.

JJ: How does it work?

GS: Not very well, but it produces some interesting results. Participants usually rely on some sort of machine learning to produce their novel, which tends to demonstrate the limitations as well as the possibilities of machine learning.

JJ: Okay, so let’s explore that. Let’s talk about what machine learning is, what kind of machines we’re talking about, and whether people are learning from the machines, or if the machines are learning from people.

GS: Well, the term “machine” is used broadly in this case. It’s like when you say, “man vs. machine.” It’s about the concept of a machine more than a specific type of machine. It’s really anything that processes data. Essentially this is about machines learning autonomously, but in practice, people have to train the machine to do that, and ultimately, they learn what the machines have discovered, so there’s learning going on all around.

JJ: But it begins with a person who teaches a machine to use data to produce some result. So once again, we see that truly transformative innovation happens when people, data, and technology are combined.

GS: Exactly, and there are three basic types of machine learning. If you just need the machine to find something specific buried in the data, that’s supervised learning. If you need the machine to look for patterns, that’s called unsupervised learning. And if you need the machine to solve a problem, that’s called reinforcement learning.

JJ: Supervised learning is great for sorting things or for decision trees. For example, suppose there are a number of possible treatments a doctor could offer a patient, each with different side effects and benefits. Even though every patient has a unique set of variables (such as age, weight, blood pressure, gender, body temperature, race, blood count, and on and on), through the use of testing sets the machine can figure out which drug would be the most effective for a given patient, while causing the fewest side effects.

GS: Unsupervised learning often involves using standard statistical modeling to find previously unrecognized associations. Using these algorithms, the machine looks for places where data is clustered to find the best possible solution. Or conversely, it can look for outliers as a way of detecting anomalies or security threats. For example, a power company could better predict hours of peak energy use based on the season, time of day, and weather by looking at a cluster model.

JJ: The third type of machine learning, reinforcement learning, is like training a dog. This is where you have a specific goal in mind, and as the machine tries to accomplish that goal, it gets either positive or negative reinforcement. You either give the dog a treat or tell him he’s a bad dog.

GS: Except machines don’t like Milk-Bones.

JJ: No, but they’re a lot more cooperative than poodles. Really, it’s like teaching a computer to play chess. Once the machine knows the rules, it can quickly run through many iterations of a wide range of variables, learning which choices are most likely to create a favorable result. But you are correct, you don’t have to tell the computer it’s a good boy.

GS: What’s really important here is the massive amount of data a machine can process very quickly. Machines have the ability to sort information, spot trends or anomalies, and iterate a multitude of variables in a way that no person could.

JJ: So why does a machine have a hard time writing a novel?

GS: Perhaps a machine will write a great novel someday. Machines only operate on the data and rules they’re given. So, let’s suppose you feed all of Ernest Hemingway’s writing into a computer. Then, using predictive models, the machine could attempt to copy his writing style by deciding the likeliest choice for each word in succession.

JJ: So, it’s just like when your text messaging app on your smartphone suggests a word.

GS: Exactly. If you could create a program that mimicked Hemingway’s writing with 100% accuracy, you would just end up with The Old Man and the Sea, or something else that he had already written. If it were perfect, you would just write the same story all over again. So terrific data alone isn’t the whole solution. You would have to have a technology that understood Hemingway’s biases, experiences, process and writing style. It would have to know how he thought, and why he rejected some ideas and used others. Right now, that’s the stuff of science fiction.

JJ: And we love science fiction. That’s why we did the Searching for Salaì podcast. It’s great to step back and imagine what might be possible one day. In fact, many software engineers are working on “deep learning,” an evolved type of machine learning.

GS: Yes, deep learning shifts from trying to imitate results, (as in the Hemingway example) to actually replicating the way the human brain thinks. This is on the path to true artificial intelligence, and they’re already seeing results such as voice and visual recognition.

JJ: This is the sort of thing that makes self-driving vehicles possible. The car knows all of the rules of the road and is able to recognize all of the objects in its surroundings. Then using reinforcement learning it’s able to get you to your destination.

GS: Now, imagine that instead of a self-driving vehicle, you have a self-driving business. The idea isn’t perfected yet, but that’s the direction this Digital Renaissance is heading. You’ll have an enterprise with technology that knows all the rules and is able to process the data in real time to help you take your business where you want it to go.

JJ: This doesn’t mean that robots are going to take over your business, any more than your self-driving car is just going to run off and do errands without you. The intelligent enterprise will have people, data, and technology working together in an extremely efficient manner.

To learn more about SAP Leonardo and machine learning, visit

For a more imaginative experience of how technology has become integrated into our lives, listen to our cool new podcast, Searching for Salaì. “Episode 5: Constructive Imagination” is available now on Apple Podcasts. Catch up on past episodes and continue the experience at

About Ginger Shimp

With more than 20 years’ experience in marketing, Ginger Shimp has been with SAP since 2004. She has won numerous awards and honors at SAP, including being designated “Top Talent” for two consecutive years. Not only is she a Professional Certified Marketer with the American Marketing Association, but she's also earned her Connoisseur's Certificate in California Reds from the Chicago Wine School. She holds a bachelor's degree in journalism from the University of San Francisco, and an MBA in marketing and managerial economics from the Kellogg Graduate School of Management at Northwestern University. Personally, Ginger is the proud mother of a precocious son and happy wife of one of YouTube's 10 EDU Gurus, Ed Shimp.

Jeff Janiszewski

About Jeff Janiszewski

Jeff Janiszewski is an SAP award-winning B2B marketeer with a proven track record of designing, implementing demand generation and pipeline strategies that generate sales across a diverse range of industries and solutions.