Admittedly, one of my all-time favorite books is “Hitchhiker’s Guide to the Galaxy.” (Let’s not talk about the 2005 movie.) It is the perfect mix of absurdist, British humor and philosophical thought. Highly recommend a read if you get bored during the great 2020 quarantine.
While there are many moments that happen in the book and related series, one of the highlights has always been the computer Deep Thought. The supercomputer is tasked with finding the meaning of life, which it does after a brisk 7.5 million years.
However, the answer made no sense because the computer couldn’t understand the question.
I always loved this moment in the story because it reminds me of the importance of intelligence over raw computing power. You can have access to a computer that can solve the mysteries of the universe, but that doesn’t mean anything unless you know what question to ask it.
Deep Thought also perfectly illustrates the importance of using machine learning properly. Machine learning parses a ton of data to produce an answer to a problem, but it might not always be a sensible answer. This is where deep learning comes into play.
Like machine learning and AI, there is a lot of confusion about what deep learning is and isn’t, so we wanted to continue our examination of these relatively new advances in data analytics.
Deep learning is often confused with machine learning and for good reason, too: deep learning is actually a subset of machine learning.
If you recall in our talk about machine learning vs. AI, machine learning is a concept that has been around for a long time. Briefly as a refresher, machine learning takes a ton of data and analyzes it, and then uses that analysis to inform future decisions.
In other words, machine learning is a basic example of a machine teaching itself.
The limitation of machine learning is that the system may come to the wrong conclusion and require human intervention to correct. In a machine learning system, even after it is corrected, it might not even understand why it was corrected. It has simply had its interpretation of the data corrected.
Deep learning, on the other hand, takes the power of machine learning and kicks it up another notch. Using layers upon layers of algorithms, deep learning parses data and tries to learn from it using multiple algorithms. Each algorithm provides a different kind of interpretation and forms what is known as an artificial neural network.
That’s basically a digital brain, and if that doesn’t sound exciting, I’m not sure what will.
For example, consider a deep learning system that examines the difference between cars and trucks. Each layer of a deep learning system will consider a different aspect of what separates a car from a truck and vice versa. For example, one algorithm might focus on whether or not the vehicle has a truck bed, while another might examine weight. Eventually, the deep learning system will create its own labels to help it parse future information.
Compare this to a machine learning system. In a machine learning system, you often have to have the data labeled or tagged for the system to sort through it. After the system has sorted through the data, it can then start understanding future data, but you still have to put in a fair amount of leg work to get the data started.
In this way, deep learning is much closer to the way that humans think. It draws on all of its inputs, considers past outcomes, and tries to accomplish its goals. Deep learning is what Netflix uses to suggest the next thing to watch, how Facebook makes suggestions on photos you might appear in, and how Spotify creates those custom playlists filled with songs.
The biggest benefit of deep learning is that you can often set it and forget it. For companies, deep learning has created a revolution in terms of performance.
For example, you can use deep learning to optimize how customers find your products on your website, assess which marketing campaigns are working most effectively, and even create automated chatbots to connect with customers.
The best part about deep learning? As it gains more and more data, it improves itself!
The most famous example of this is Google AlphaGo, which learned how to play the incredibly complicated game of Go all on its own. Eventually, it even became good enough to beat grandmasters!
If you’re interested in learning how to implement deep learning at your company, contact Hoverstate today. It doesn’t matter what industry you’re a part of; you have problems that can be solved by the power of deep learning. We’ll not just help you come up with solutions–we’ll help you figure out the questions that deep learning can solve.