Machine Learning: Why Now?

Machine Learning: Why Now?

In a 2016 Google technical conference keynote, Eric Schmidt, who is now chairman of Google’s parent Alphabet, said he has seen the future of wealth creation from the IT industry, and its name is machine learning. And earlier this year, in a Newsweek opinion piece, Schmidt wrote about the significant impact he sees machine learning having on the world. He said, “We are now entering a decade in which machine learning will come to define how we interact with technology and the world around us – and how technology helps humanity thrive.”

I agree with Schmidt’s assessment and believe that digital technologies such as machine learning have the power to change how the world creates innovative value. This powerful technology is bringing companies unprecedented insights, more accurate predictions, and the automation of routine tasks, all of which allows them to focus on higher value opportunities.

Problem-solving beyond human comprehension

No longer are programmers defining rules for computers – with machine learning, we now give them problems to solve and they learn how to accomplish this on their own. Imagine software that can observe human actions and learn how to perform them autonomously without being explicitly programmed. Imagine machines that can access, analyze, and find patterns in Big Data that is beyond human understanding. Imagine machines that can spot patterns and make connections through exposure to massive amounts of data so they become intelligent advisors and help people make better decisions.

Machine learning is doing all of this and it’s helping companies become exponential – where they have an impact that is at least 10 times greater than their competitors.’ And it’s creating intelligent enterprises where businesses are more empowered than ever before.

Why all the hoopla—and why now?

You may wonder why there is so much in the news about machine learning now. Why now, when artificial intelligence, the parent technology to machine learning, has been around for more than 50 years? The reason is because there is an extraordinary convergence of large volumes of Big Data, unprecedented computing power, and sophisticated self-learning algorithms taking place. The affordability, viability, and feasibility of these three technologies are the driving forces behind why machine learning is becoming more and more prevalent today.

Let’s look at Big Data first.

The amount of data available today – we are currently creating around five exabytes a day roughly equivalent to 500 million songs – is giving machines the possibility to become super-intelligent. Nonetheless, the amount of computing power required to process all this data takes far more computing power than that available in everyday central processing units (CPUs).

Not long ago, researchers found that instead of stacking traditional computers with CPUs in the cloud, another approach could be more effective. The graphics processing units (GPUs), which were developed to speed up the graphics in our everyday computers and especially in the gaming world, had faced a similar problem: the growing sophistication, data volume, and data speed in computer graphics. They solved the challenge by stacking a very large number of simpler computing units that would work in parallel to render the graphics.

Researchers realized the same approach could be a way to accelerate ML computations, and their talent made it work. As an indication of the power unleashed, I’ve been told that it would take 2,000 CPUs to match the power of 10 GPUs. With the ability to execute in-memory data management and their tremendous parallel processing capabilities, GPUs have given machine learning a good part of its current thrust.

If you now add to this the network effect provided by the open availability of highly sophisticated shared algorithms for ML, it’s should be easier to understand why now machines can think more accurately and independently, recognize contexts, and make better decisions on their own. What these machines will do for us is already going well beyond what predictive analytics and Big Data analytics have done for us in the past – and well beyond what any human can do in may domains!

Machine learning in the enterprise

For enterprises, machine learning can automate and prioritize routine decisions, making processes leaner and faster. Machine learning can also change traditional rules-based processes into intelligent processes by discovering and exploiting new patterns in large, unstructured data sets, making predictions about them, and adapting the processes accordingly.

In practical terms, what does this all mean? Well, for instance, when machine learning is coupled with technologies such as the Internet of Things, a machine can decide on what’s optimal to fix in the first place in a manufacturing plant. In a human resources department, machines can intelligently match resumes to open positions and recommend career paths by matching skills to future enterprise needs.

In finance, machine learning can take over lightweight finance operations and some of its highly repetitive tasks, such as checking inter-company reconciliations, invoices, or travel expenses for accuracy. It can as well interact with suppliers and provide optimal procurement recommendations to the purchasing department. Machine learning can also facilitate self-driving customer service where tickets can be created intelligently—i.e., they are created if ML thinks they are needed, even if the customer did not open them—and they can be handled by the customer care personnel guided by accurate step-by-step recommendations from the ML system.

This technology is clearly transforming applications throughout the enterprise – at exponential speed.

Author : Felipe Gomez
Degree : Masters
Major : Marketing and Advertising
Country : United States
Language : English

View Resume