October 7, 2019 - Machine learning is an emerging element within the field of artificial intelligence that allows a computer to, in a very real sense, teach itself. It learns how to solve problems without being taught explicitly by humans how to do it.
There are several common machine learning techniques. In a nutshell, a computer program using machine learning can determine a solution to a problem by analyzing data on its own, then self-correcting by comparing its decisions and assumptions to real-world data.
In this way, artificial intelligence (AI, for short) can learn how to win a computer game, operate a self-driving car, mine ore deposits, or create the optimal delivery schedule for a manufacturing company.
Never sleeps and sees everything
A fast-growing field within machine learning is exposing computers to data captured through cameras and image analytics. This has a lot of practical business applications because it allows operations to accelerate as machines rapidly learn about a task visually.
Humans, for example, might adjust their approach to a problem incrementally, over a span of days, weeks, or even months, first collecting data and later producing reports, recommending courses of action, and then reviewing results, which feeds back into the decision loop.
But AI has the ability to make many adjustments in real time or near-real time, finding the optimal solution in a fraction of the time. An AI solution never sleeps and sees everything the camera and other data streams are capable of capturing. It doesn’t miss details through inattention or mistakes.
New industry opportunities
We can see this taking root in many industries. In the construction industry, for example, where safety is paramount and accidents happen because of inattention to wearing personal protective equipment, there are active pilot programs in which AI technology is being trained to monitor the worksite to make sure everyone is properly equipped with protective equipment.
And we’re just getting started.
Consider digital personal assistants and voice recognition technology, for example. It’s also on the cusp of a major transformation thanks to advances in machine learning.
Right now, voice system technologies aren’t especially smart, and so automated call centers, for example, have very little flexibility to react to a caller in a natural, organic, or human way.
For the most part, voice system technologies make strictly tree based decisions, offering options to callers based on a limited set of choices. But a smart voice system could parse keywords from the call, homing in on the purpose of the call asked with natural language.
Even the tone of the caller could help inform the AI on how to react, as it learns how to interpret tone of voice and how certain phrases inform how the caller is feeling.
In the short term, that means being better able to hand off the caller to someone who’s equipped to handle their specific needs; in the long term, the automated system itself could help solve the problem while remaining sensitive to the caller’s temperament.
Making robots ‘smart’
And then there’s robotics – another fast-changing field. We’ve had robots in industry for decades, but the robots we’re traditionally familiar with have no flexibility or nuance built in.
The robots perform the same functions repeatedly, without exception. In fact, contemporary robots are built for repetition, often with wide yellow caution lines painted permanently into the floor around them to prevent workplace injuries, since the robots only know their job, and not the presence of people around them.
But that’s going to change. Thanks to machine learning, robots will get smarter and more flexible, able to use cameras, real time data about their surroundings, and data about the state of the task in front of it to work more efficiently, while also reacting to humans within its “personal space.”
First and foremost, that will help robots avoid injuring their human co-workers, but later it will also improve how the robots work by interacting with and responding to the humans in their workflow.
The growing role of executives
Obviously, AI is a field in the midst of a major transformation. Thanks to machine learning, modern AI is getting phenomenally smart, increasingly able to do everything from medical diagnosis to driving cars to predictive maintenance on assembly lines.
Executives driving AI and machine learning projects will need to pursue several tasks in order to successfully adapt this transformational technology:
- Set the stage within the organization to encourage the kind of collaboration and experimentation that’ll be needed to pilot these technologies.
- Have the vision to know what the right use cases are within their industry.
- Have a plan to smartly handle all that data that’s going to come in.
- Ensure compliance with legal guidelines around handling data that may well include information about customers and the public.
It’s a tall order, but the potential benefits to business are almost incalculable.