Acquiring Machine Learning is Not Hard; It’s About Aligning Skills

While there is a certain mystery that is associated with machine learning, in some ways, these concepts are not entirely new. 

Over the past few years, we’ve seen a dramatic increase in discussions around emerging technologies like AI (artificial intelligence). At any conference you go to, whether it be procurement, finance or any other business discipline or industry, words like machine learning and AI are thrown around all over the place and catch fire to anyone willing to listen.

While there is a certain mystery that is associated with AI, in many ways these technologies are not new. In fact, the evolution of machine learning started as early as 1950, when Alan Turing created the world-famous Turing Test. It stated that “for a computer to pass, it has to be able to convince a human that it is a human and not a computer.” Since then great advances have been made in all areas of AI, with one of the most recent ones being the Google AlphaGo algorithm that won the Chinese Board Game Go competition in 2016.

But one might wonder, why now? Why has interest in this field suddenly accelerated and advanced outside of the lab and into business? According to Professor Jaakkola at the MIT Computer Science and Artificial Intelligence Laboratory, there are four main reasons for the recent advances made in machine learning, as follows:

1. The accumulation of huge amounts of data – sometimes referred to as Big Data
2. Advances in computational power – consider Moore’s law
3. The growing complexity of data models
4. New possibilities created by deep learning architectures

Moreover, while all these statements are true, I would postulate that beyond what Prof Jaakkola states, probably one of the biggest reasons we’ve seen a rapid increase in interest is based on the commercialization of machine learning tools and accessibility to them. If you look around, today’s IT organizations have access to several data sciences and machine learning platforms, including Microsoft’s Azure Machine Learning studio or TIBCO Software, recognized brand names for decades in the field of enterprise computing.

Therefore, from a business perspective, the acquisition of machine learning techniques is not the hard part. For example, organizations globally are using proven supervised learning algorithms like k-nearest neighbors, linear regression, and/or naive Bayes to improve pattern recognition and help predict business outcomes. In other cases, organizations are looking towards advancements in unsupervised machine learning algorithms that can solve problems by detecting new relationships and interesting patterns in completely new data without pre-existing models to learn or train from – think new ways to invest or approaches to curing cancer. Still, others are looking to leverage, essentially programs that look to find the best possible behavior or path it should take in a specific situation – consider sensor readings where the algorithm must choose the robot’s next action.

No question, as AI technologies become more advanced, machine learning algorithms will continue to make daily life easier by automating tasks and sorting through vast amounts of data with a level of speed and data accuracy that is superior to that of humans in the ability to make new connections between data to enhance knowledge capture for actionable insights.

Moreover, as these techniques are applied to a wide range of business problems to deliver tangible business value, it will continue to transform how people live and work. This question of job replacement is a real concern for how AI adoption will take place and the change in the workforce. For instance, according to McKinsey, job profiles characterized by repetitive activities or that require a low level of digital skills could experience the largest decline as a share of total employment to around 30 percent by 2030, from some 40 percent. The largest gain in share could be in non-repetitive activities and those that require high digital skills, rising from roughly 40 percent to more than 50 percent.

Hence, to run and manage these AI processes, these high digital skills like data science disciplines will be needed to design, understand and use computer programming that learns from experiences to improving classification, modeling and prediction. Companies will need to take on the task of acquiring skillset or retraining  people to work with AI. But just like any other technology, a business perspective is also needed. While you may have all the tools in the war chest and have the scientists trained on them, knowing how to use them properly also comes with experience and intuition in understanding the underlying problem, something that can’t be easily replaced or learned. Trying to fix a problem with technology alone, with machine learning or other AI, could be a fool’s errand without having deep experience in the skills that you are trying to automate.

Applying machine learning in procurement areas like spend analysis has important consequences for improving both the classification of spend data and for predictive insights from it, human intuition cannot be replaced by machines alone, at least not yet. Thus, from a procurement perspective, and more specifically a spend analysis perspective from my role at SpendHQ, organizations in all industries should look to accelerate the adoption of machine learning and other AI disciplines (e.g., NLP, neural networks, RPA) in concert with digital transformation efforts, but leverage years of experience in handling spend data by keeping a partnering focus in mind.

To learn more about the topic of machine learning and its applications for procurement, join us for an interesting discussion on Thursday, March 28th for our webinar, Machine Learning and Its Impact on Spend Analysis.