To ignore the benefits of deep learning would be to accept watching your competitors steal a march. But how does it work and what are the benefits?
Data is valuable. We all know that. But the real question is: How much do you believe it?
Answering that question will guide your approach and commitment to your investment in deep learning. In this case, belief really means conviction. Do you believe your data is full of valuable information the way you believe your mother loves you (viscerally, no questions asked!), or do you believe it the way you believe your dog loves you - skeptically. He might love you, or he might just remember that you’re the person who feeds him. Or maybe a bit of both.
Ignoring deep learning today is like ignoring spreadsheets in the early 80s
If your conviction about the value of your data reaches maternal depths, you should be on your way to exploring deep learning techniques. If your conviction is more canine, you’re not ready to jump in with both feet, so don’t let anyone convince to do it! Still, pay attention. Experiment. Learn.
Ignoring deep learning today is like ignoring spreadsheets in the early 80s. You could, and get away with it for a while, but eventually more nimble competitors used spreadsheets to manipulate data so fast, they crushed competitors who insisted that paper and pencil were still good enough. Deep learning is in that category.
What is deep learning?
Let’s take a quick step back.
Deep learning is an umbrella term for a class of algorithms that allow us to “teach” a computer to detect, categorise and apply patterns in data. What’s interesting about these algorithms is that they’re very fuzzy. This is handy when we want to find a pattern, or the absence of a pattern.
For example, we might use deep learning to teach a computer to recognize umbrellas. To do that we’ll gather up thousands of photos of umbrella to teach the computer what an umbrella looks like. The deep learning algorithms won’t just learn to recognize those umbrellas, but, in theory, any umbrella. Pretty nifty.
Another example is the detection of fraud. We can use thousands of data points that represent valid credit card transactions, and then let the computer evaluate future transactions against what it’s learned. In this case, we aren’t interested to know which transaction matches our training set, but rather, those that don’t. They are most likely to be fraudulent.
If by this point you’re thinking “Wow, that sounds amazing, but how does that relate to my business?” Listen up.
No matter what kind of business it is, your business produces data. Every time you interact with an employee, a customer or supplier, data is created. Whether or not you adequately collect that data is the subject of another post. For now, we’re going to assume you do.
Now, what if you could use that data to make predictions about the future? What if you could collect enough data during an interview to predict how likely a job candidate was going to make a good employee? What if you could determine after a first interaction whether a customer was going to be a good customer long-term, or even just a good credit risk? Any of those would make your business more efficient, less risky, and far more competitive. Deep learning techniques could help you answer all those questions.
Not a question of 'if' but 'when'
Investing in deep learning is necessary. You may be able to get away with ignoring it for a few more years, but you do at your own risk. Eventually you’ll be like the baker who spent hours updating his paper ledger while his competitor handled it in minutes with a spreadsheet. You might be good at synthesizing and evaluating a lot of fuzzy data, but if the data’s good enough, you probably can’t beat a dispassionate machine. Already, deep learning computer vision algorithms are outperforming human beings. It won’t be long before they’re helping business across the globe run better than we can.
Paul Murphy will be speaking at the RE.WORK Deep Learning Summit in London on September 24th & 25th. Virgin readers can save 20 per cent off all ticket types using discount code VirginEnt. Register before July 31st and save on the already discounted Early Bird pass.