Remember when “big data” was the sexiest term at a strategy off-site? The good old days of dashboards, charts, and executives pretending to understand regression analysis. Now the new, new corporate crush is machine learning. The board wants to know when the business will adopt or managers want to be the first to take it to the market.
The promise is intoxicating:
- Algorithms learn,
- Systems predict,
- Machines recognize faces, voices, patterns, and
- your preference for whole milk or skim, before you do.
Somewhere between “Hey Siri, what’s the Klingon word for: rude” and “Alexa, order more toner,” an idea took hold that machines can learn faster than people, more reliably than people, and with less demand for a 401(k) match that people might expect.
Machine Learning
Machine Learning (ML) is a branch of computer science that allow systems to detect patterns or make predictions from data rather than being explicitly programmed for every scenario the machine learns. For a machine to learn, people will design an algorithm to use historical examples to build a model and then the model can apply to new situations.
Machine learning sits inside the larger artificial intelligence umbrella, though jargon can blur the view. Artificial intelligence, or AI, pursues ways to make machines act with a kind of reasoning reserved for humans.
Business seems to believe that machine learning is the way to get there. Instead of programming every rule, people feed machines data so machines recognize patterns and make their own predictions.
Welcome to the Machine
ML can automate tasks that involve recognizing patterns such as fraud detection in banking, forecast sales, demand, risk outcomes, and personalize experiences and recommendations such as in retail. By doing so, it can reduce manual workload, improve accuracy, speed decision-making, and uncover insights hidden in large data volumes.
But this affair with machine learning comes with a blind spot that assumes organizations know what they teach machines in the first place.
Machine learning for relative context includes Amazon or Netfilx recommendations.
Speed Kills
A bad decision process plus feed into a sophisticated model equals an expensive way to speed up bad decisions. Automating poor thinking creates bad results faster.
As the late W. Edwards Deming might say:
“If you cannot describe what you are doing as a process, you do not know what you are doing.”
Try sifting that through your neural net.
Machine learning succeeds where data reflects disciplined thinking. But many organizations treat data as a mirror for wishful thinking. Garbage in = neural garbage out.
Early adopters such as: Amazon, Google, and Netflix have built their cultures around experimentation and error tolerance.
Their reward?
Feedback loops that refine themselves.
For everyone else, the temptation is to plug machine learning into cultures that punish mistakes. That is like planting a rainforest in a parking lot.
Machine Learning Thrives on Learning
Bureaucracy thrives on control. Worse: people have bias. Machines are not compatible biological systems where motivation is a competitive advantage.
In a recent article from The Economist, Machine-learning promises to shake up large swathes of finance, expected Machine Learning improvement includes:
- Trading,
- Credit assessment, and
- Fraud prevention
No matter the basis, I smell bias.
The more exciting question is not what machine learning will learn, but what it will unlearn from us. As I noted in The cost of culture, a 50 % turnover of the Fortune 500 you cannot bolt machine learning onto an organization without cultural readiness.
In first-mover zeal, what companies will hand over operational judgment to algorithms and surrender the very thing that makes human capital valuable: discernment. The next big differentiator in organizations will not be who has the smartest machine, but who still has people capable of asking why.
Machine Motivation
Curiosity is king. Compare the impact of continuous learning between people and machines on data privacy.
Machine learning reshapes every process that depends on prediction. But leadership will remain the process that defines priority.
As data gets smarter, the human side of organizations must get wiser or risk becoming the dullest part of the system.
Having a crush does not always mean love. In business, falling in love is an emotion that many leaders can’t afford.
Are you asking your organization what your machines are learning and more importantly what they are unlearning from the people?