Cool Technology and the Intelligent Enterprise
A Five-Minute Survey
Yo! You guys are not gonna believe the guy I met yesterday. I know we were all told never talk to strangers, but he wasn’t that strange after all, and he also had a lot to say about cool technology we use every day of our lives whether we’re aware or not. And like I said before, if you’re new to this blog, you might wanna catch up here. Keep reading, my friends.
I climbed the steps to the Dempster L, scanned my Ventra card, and headed down the platform. There was only one other guy waiting there, but sure enough, as soon as I was spotted, he headed right for me. Naturally, I felt a hustle coming on. That’s life in the big city. You begin to expect it. So I turned my head to avoid eye contact, but he was undeterred. “Excuse me,” he called, “Can I ask you a few survey questions for a survey?”
Suddenly my brain froze as I tried to decide whether that sentence was a redundancy or just awkward. Meanwhile as my head tried to reboot, my mouth defaulted to, “Sure.” An utterance I thought I would regret.
“My name is Norman Powell, and I work for the Chicago Transit Authority,” he babbled rapid-fire brandishing a badge that was strung around his neck. “We’re doing a short ridership survey to learn how we can improve our ridership. It will only take five minutes of your time and we can conduct the survey while you ride the train because it will only take five minutes of your time. Would you like to participate?”
“You bet, Norman. Anything to help.”
“Yes. It’s just a five-minute survey, and your responses will help provide the best customer service possible.”
“You’re very focused Norman. Fire away.”
“Great. We’re not permitted to collect any specific identifiable personal information, but we would like some basic demographic information.” He then proceeded to ask about my age, gender, and race. He asked what neighborhood I lived in, how long I’d lived there, and how often I take the train.
I cooperated without inserting my usual wisecracks, but finally, I said, “Come on Norman, you guys got all that information when I signed up for the Ventra card. And when I swiped it tonight, you knew exactly where I traveled and how long I stayed there.”
“While we do collect aggregate data, we’re not permitted to collect any specific identifiable personal information.”
“Yeah, I got that part, but don’t you have any harder questions for me? The square root of two? The capital of Wyoming? What I got for my 12th birthday?”
“That would be specific identifiable personal information.”
“Okay. Sorry. You’re right.”
The game was on. Norman wanted to fill out the survey as efficiently as possible, and I wanted to throw him off stride and have a genuine conversation. I love playing that game. I got the sense Norman was an interesting guy, and I wanted to know what made him tick.
“We would like to know where you went tonight, how long you were there, and whether it was for business or pleasure — keeping in mind that you don’t need to give us, nor do we want, any specific identifiable personal information.”
“Yeah. You have any kids, Norman?” He looked stumped, like no one ever asked him a question before. “It’s a rhetorical question, Norman. I don’t need your personal information and you don’t want mine. Got it. The point is that I don’t have any kids, but I just spent the last three hours at a youth soccer match, and now that I think about it, it sounds a little creepy.”
“We’re not permitted to collect any specific identifiable . . .”
“Yeah, well I was also interviewing a woman about data intelligence for this blog I write. You know what data intelligence is?”
His eyes lit up. “That’s what I do for the CTA. We use this survey to verify and supplement our machine learning.”
Ah ha! I’d knocked him off his game, but now I was intrigued. “Machine learning? Like robots or something?”
“No. Not at all. It has nothing to do with robots. You see there are all of these interconnected things out there collecting mountains of data.”
“The Internet of Things. I know about that.”
“And obviously you know what data intelligence is, right? Well, data analysts manage, sort, and represent data in a meaningful way, but machine learning is about leveraging the data autonomously.”
“But not with a robot. What kind of machine are we talking about?”
“When we say ‘machine’ we’re talking about some kind of computer.”
“And who’s doing the learning? Are people learning from the computers or are computers learning from people?”
“Well, to be precise, people tell the computers what and how to learn, and the computers report back what they’ve discovered. There are three basic kinds of machine learning: supervised learning, unsupervised learning, and reinforcement learning. The CTA uses all three. Supervised learning is when a computer is instructed to look for a specific result. The best example is a spam filter. First, the computer looks for the frequency of certain words found in known spam emails.”
“Like ‘Nigerian miracle cure’?”
“Exactly. Then the computer will build a rule to kick out any email with those words.”
“So, the machine learns how to filter spam.”
“Exactly. Unsupervised learning applies common statistical methods and algorithms to a data set, and then looks for clusters of results to discover previously unknown relationships. This is how medical researchers discover relationships with drug interactions or disease —” He was momentarily interrupted by the roar of the train and our conversation halted. We both stepped onto the empty car and, tossing aside some discarded newspapers, we took seats near the door.
As the train pulled away, he picked up where he’d left off. “It’s how researchers discover disease-causing agents, or how actuaries identify risks. The third kind of machine learning is called reinforcement learning. It’s an iterative process. The computer is given a goal and lots of data, and through a complex trial-and-error process it decides whether the inclusion of certain data gets you closer to or further from the goal. It’s like playing a really complicated game of chess.”
“So, you play games for a living?”
“Sort of. The CTA used to operate on anecdotes, tribal knowledge, and gut-driven decisions, but that was crazy. On a typical weekday, we have more than 3,000 busses and rail cars taking 1.6 million riders on more than 20,000 trips. It’s very complicated. There’re a lot of decisions to make, and in today’s world there’s no such thing as a good decision made without data.”
“So, you’re not using robots, but you are replacing people with machines.”
“Not at all. We’re empowering people with knowledge that is so complex it can only be extracted by a machine.”
“So why are you taking a survey? Aren’t you going back to anectodical evidence?”
“Just double-checking to make sure we got it right. This train is almost completely empty, right? But in about 15 minutes we’re going to pull into the Addison station just as the Cubs game lets out, and this train will be packed. I’ve been watching the progress on my tablet, and unless Milwaukee ties it up in the top of the ninth, we got it right.”
“I love the Cubs, hate the Brewers. But what’s your batting average? You always get it right?”
“No, but we get it right most of the time, and we’re significantly better than we used to be. There’re a lot of variables out there. We begin by looking at cycle time — how long it takes a train to complete its route and turn around. We look at track occupancy data and signal data, and we consider station congestion at transfer points. Then we look at maintenance and human resources. We look at car length, operator experience, passenger dwell time, the likelihood of an accident, and construction and slow zones. We also consider external data like weather, parts availability, holidays, and special events.”
“So, the more data you include the better the result?”
“Not at all. It still requires the data to be weighted. For example, what has a greater effect, operator experience or weather? Perhaps those variables are completely irrelevant, and their inclusion would disrupt the accuracy of the model. And some things are almost completely unpredictable, like the Cubs’ World Series win in 2016. We knew Wrigleyville would go nuts and the impact on the CTA would be huge. But we didn’t know if they would win in Game 4, 5, 6, or 7, or maybe just lose the series. As it turned out, they won in Game 7, in extra innings, after a rain delay. You can’t predict that, but you can expect and plan for the unexpected.”
“It sounds incredibly complicated.”
“Well, the concept is quite sophisticated, but the goal is to have the simplest, most effective model possible. And because machine learning is truly a learning process, we get better and better at it every day.”
“There we go again. That concept has been haunting me for days. It seems intuitive that if you could simplify everything, you could do anything. But the act of simplification is always very complicated.”
As we approached the Addison stop, I finished Norman’s survey, and he thanked me for my time. Then he plunged from the train into the Wrigleyville crowds, while I stayed on to the Belmont stop. The Cubs did win, and the train was packed. When I finally pushed my way onto the Belmont platform, a familiar image appeared before me. It was da Vinci’s Vitruvian Man on a billboard, and at the bottom it read, “Join the SAPien movement — August 18 at McCormick Place.” Now I knew it was more than fate; it was a movement.
But what could it mean?
As I exited the station, I put on my earbuds and listened to Charlotte’s podcast about her time-traveling friend, Jack “Salaì” Finney. I don’t know what to make of the guy, and I guess she doesn’t either. You’ll just have to take a listen to Charlotte’s show on Apple Podcast, and judge for yourself. In my next blog I meet a very interesting guy down by “the bean” and I learn that “the cloud” is no laughing matter.
###
Tony
Blogger Extraordinaire
DISCLAIMER: The Searching for Salaì podcast and the “Cool Technology and the Intelligent Enterprise” blog series are works of fiction. Names, characters, businesses, places, events, locales, and incidents are either the products of the author’s imagination or used in a fictitious manner. Any resemblance to actual persons, living or dead, or actual events is purely coincidental.
[DESCRIPTION]
Tony answers a survey and learns that machine learning is a “learning” process, and thus we get better and better using it in our everyday lives every day.
[KEYWORDS]
SAP Leonardo podcast, Innovation, Digital Transformation, Intelligent Enterprise, Analytics, Big Data, Blockchain, The Cloud, Design Thinking, The Internet of Things (IoT), Machine Learning, Data Intelligence, Digital Renaissance, Leonardo da Vinci, Searching for Salaì, Cool Technology
Comments
Post a Comment