A Deeper Dive on Deep Learning

Lukasa Insights

Listen to article

Leveraging the best of AI for actionable insights 

“The human brain has 100 billion neurons, each neuron connected to 10 thousand other neurons. Sitting on your shoulders is the most complicated object in the known universe.” Michio Kaku, physicist

No computer can match the abilities of the majestic human brain. A computer can’t imagine, dream, create, feel, experience, philosophize, or ponder its own existence—it can’t think, in the truest sense.

And yet, there are functions that a computer can perform with far greater speed and efficiency than the human brain. 

Deep learning is a subfield of machine learning (ML) and artificial intelligence (AI) that is modeled after the structure of the human brain, and attempts to imitate the way it functions.

Deep learning is based on a neural network—brain-like artificial neurons connected by synapses—forming multiple layers of processing that progressively teach the computer to recognize patterns and identify, label, and categorize abstract objects. Simply speaking, it is a sophisticated form of predictive analytics, in which each new level of learning is built on, and refines, the one before it. The more data the computer is fed, the better and more accurate its performance becomes. 

The accuracy of deep learning, says Forbes, “has led to the greatest leap in the history of AI. Today, nearly all state-of-the-art AI is based on deep learning.”

Although it is a subset of machine learning, deep learning is distinct in two main ways: the type of data it works with, and how it learns from that data. 

In classical machine learning, a computer makes predictions based on structured, labeled data (chiefly numeric) that it has organized into tables and relational models. However, a human has predefined the features by which the computer sorts the data.

Deep learning, by contrast, allows a computer to take in vast quantities of raw, unstructured data (diverse types of information that do not fall neatly into predetermined models, including photo/video/audio/text/etc.) and, on its own, without the input of a human expert, create and define a hierarchy of properties by which to sort it. The computer makes decisions and predictions based on patterns it detects, constantly honing and reinforcing its learning with subsequent data. 

This is an extremely useful tool for data scientists because despite the human brain’s incredible ability to recognize patterns and learn from them—making decisions and predictions and clustering information based on the raw data we take in—humans aren’t very adept at understanding and articulating the properties by which our minds actually sort the data. This narrows our ability to define the features we should instruct a computer to sort data by.

To use a common example, even young children can tell the difference between cats and dogs. But why and how? If you ask someone to tell you what properties distinguish each species, chances are they can only identify a few. Cats and dogs actually share many key features, so we’re hard pressed to name offhand all the criteria our brains must be using to determine which is which. 

With deep learning however, a computer is not dependent on a human telling it what criteria to sort by, and therefore not limited by our ability to extract features of the data. It uses all the available data, extracting criteria as it goes, getting “smarter”—more and more specific and accurate—with each layer of processing and every new influx of data it receives.

This may all sound intimidatingly high-tech with a splash of sci-fi. But while we tend to lump it all under one generic, mysterious label of “AI,” the reality is that deep learning, specifically, is already very present in our daily lives—every time you use a digital assistant or voice control in a device, receive an automated personalized suggestion of what product to buy, or get a credit card fraud alert. It’s also powering emerging technologies like self-driving cars; helping doctors detect diseases and sports teams optimize player performance. 

Deep learning has very practical applications and can provide straightforward, actionable insights. It has enormous value for individual businesses and entire industries. Deep learning can help you increase productivity by automating repetitive tasks, reduce errors, enhance your product, improve workplace safety, know your customer better and serve them faster and more effectively, detect risks and anomalies, identify opportunities, and much more—ultimately saving your company money and driving revenue.

The business and technology experts at Lukasa specialize in process analysis, modernization, and digital transformation. We take a partnership approach to every project, working side-by-side with your team to help you identify and seamlessly integrate the most effective AI, ML, and deep learning solutions for your unique enterprise. 

About Lukasa - lukasa.com
Lukasa empowers small-to-medium-sized businesses by designing and implementing custom business and technology solutions that drive efficiency, productivity, and innovation, enabling them to stay ahead in today’s rapidly-changing competitive landscape.