Sergey Nivens - Fotolia
Looking for something else?
As I write this article I am sitting on a really long flight. I have flown so much in the past six weeks that I have gone through every movie the airlines have offered up. I am now so desperate for flight-time entertainment that I am watching movies from the "classics" list. Right now I am watching Moneyball. Now, that might not sound like much of a classic, but for a data nerd like me it fits the bill. Watching Moneyball got me thinking about how much advanced analytics have evolved and changed over the years. At the time (2002), what the Oakland A's did was revolutionary. Now it is commonplace.
What is revolutionary is how machine learning, artificial intelligence and cognitive systems are advancing and changing our world. What is revolutionary is how much more data we have to feed into our data modeling and cognitive systems. I am guessing that if Moneyball were remade today, the number crunchers would include players' Facebook data, their credit card and spending data and the data gleaned from their grocery store purchases as recorded by their loyalty systems. The data picture would be much more encompassing but also much more complex.
And that is where machine learning, artificial intelligence and cognitive systems come in. The advances and changes in these technologies and approaches are staggering. Perhaps it is time for all of us to experiment with and then start using machine learning, artificial intelligence and cognitive systems.
But first, as a longtime data nerd, I will force you to endure my short version of the history of these really interesting topics.
For a long time now people have been doing research in two broad areas: One, thinking/learning systems and, two, data mapping, modeling and analysis methods. This research was pretty esoteric and highly specialized but not really consumable by anyone but the researchers because both areas required massive amounts of compute power in order to do anything useful -- and that massive compute power was too expensive for anyone except those in large research institutions to do and use.
Well, the cloud changed the economics of massive compute. Instead of purchasing the massive compute power needed to do meaningful machine learning, artificial intelligence, cognitive systems and the data mapping using those esoteric models, someone could rent a massive number of compute nodes for a few minutes, hours or days and then turn it off once the work was done. Even better, once that massive and now inexpensive compute power was available, other service providers leveraged that to build cloud services to deliver on-demand services such as text analytics, statistical modeling, machine learning algorithms and pretty much anything you might need -- as long as you have the data all those need.
What we're doing with AI, ML
With the barriers to usage so low, everyone has jumped on board. It seems that pretty much every vendor now includes the words "machine learning" in their sales pitch and there are new data analysis technologies and companies showing up all the time. Now, it could be that these vendors and technologies are high on promise and low on delivery, but to me this shows that pretty much everyone can do something with machine learning, artificial intelligence or cognitive systems if they want to. And, from my perspective, the sooner the better. Perhaps it would help if I explained some of the things we are doing to be a bit on the leading edge.
- Our business is about connections. Our clients hire us to help them improve their employee engagement. We provide a number of products and services, but they all depend on connecting employees to employees, managers to teams and executive leadership to everyone. Well, sometimes connections are not obvious. How can we sort out who interacts with whom? It turns out that we don't need to sort that out. All we have to do is condition our data and send it to a cloud social-graphing service and let the service do its thing.
- Our business is also about a whole range of cause/effect relationships. How do we optimize a marketing initiative? How do we optimize our inventory management and our product merchandising? Well, we don't. Instead we let machine learning algorithms propose the optimal models, because crunching massive amounts of data and finding cause/effect relationships is what machine learning does best.
- Our business is also about gaining competitive advantage in the marketplace and one of the ways we are doing that is by productizing our data. This means we use data analyses to identify key insights that we then provide to customers. These insights are critical enough that it cements our relationship with our customers or are compelling enough that it creates reasons for our prospective customers to choose us.
So, there is your assignment. Pick one opportunity -- some cause/effect relationship that will improve your operations or marketing or products and services or some key insight you can gain or some new approach to your products and services -- and do a machine learning, artificial intelligence or cognitive systems experiment. Take a flyer. Do something bold. Do something a bit risky. Try out chatbots. Do some text analytics. Play with social graphs. The costs are low enough that even if you fail, you have lost very little and perhaps gained valuable experience -- experience you will take into your next experiment.
As in every other technology wave, something compelling will survive the hype and when that happens, I want to be fluent in its use. And let me know how the assignment goes. Shoot me an email at NNick@octanner.com.
About the author:
Niel Nickolaisen is a veteran IT leader, currently serving as the CTO at O.C. Tanner Co. Niel is a frequent writer and speaker on transforming IT and on IT leadership.
Recent advice from Nickolaisen:
Finding IT agility in the OpenStack platform
Seven tenets to live by when adopting DevOps