Our Blog

“Knowledge is knowing a tomato is a fruit. Wisdom is not putting it in a fruit salad.”

Categories

  • Categories

Search Blog


What is deep learning?


Artificial intelligence encompasses a lot of different things; perhaps one of the coolest is deep learning.

 

So what is deep learning?

Within AI, there’s machine learning, and deep learning is a subset of that. Both have been around for a long time, so while it sounds very cutting-edge, it’s also very proven, powerful technology! In fact, deep learning has shattered many long-standing performance benchmarks.

So to start with machine learning, this is where you feed data into a machine so that the machine can learn to recognize certain things. Here’s a quick explanation of how that works from our Chief Scientist, Dr. Neil Yager:

 

 

In this example, we talk about facial recognition software, which is one branch of AI. To teach a machine how to recognize faces, you’d feed in a dataset of thousands of images. Some of them would have faces, and you’d tag them as such, and some would not, and you’d tag those accordingly. After the machine has enough “training”, you could give it an image without a tag and have the machine assign the tag of “face” or “not face” to the image with a pretty good degree of accuracy.

 

Deep learning takes this a step further, and the idea is that the more data you have, the better the machine will be able to predict what happens next.

Deep learning uses something called neural networks. These networks consist of multiple layers of operations, so the “deep” in deep learning is referring to those additional layers. Neural networks allow deep learning models to not need as much hands-on management by a human. They can decide which features to base decisions on, and identify the mistakes they make and then learn from them.

What makes deep learning so powerful is the huge amount of data it can process.

It’s now possible to scrape the entire internet to make a gigantic dataset that you can use to train natural language processing algorithms. Previous machine learning techniques couldn’t handle this much data.

Image credit: Disney
via GIPHY

 

One example of deep learning is Siri, which has to learn your individual voice.

Siri’s been trained on lots of voices and accents, so this shouldn’t be a problem, but you may notice that the first few times you use Siri, your results might not be quite as expected.

But the more you use Siri – the more “she” hears your voice and interacts with you – the more data you’re providing, and the decisions that Siri makes on how to help with your requests gets smarter.

 

Phrasee is another example of technology that uses deep learning.

We start out by taking into account your dataset (i.e. your marketing communications) plus any other available tone of voice documents, and then use natural language generation (NLG), another type of AI, to generate language that’s appropriate to your brand and optimized to drive the best results.

The first time, it’s pretty good! But the results get even better over time as we gather more data and make better predictions about what’s going to resonate best with your audience. The deep learning comes in when we start to get data from your split tests to improve our decision-making.

Image credit: CBS Studios
via GIPHY

 

So can deep learning solve all of our problems? Not exactly.

As with all machine learning, it is only as good as the data it is trained on. Also, there are many problems that aren’t a good fit for deep learning, such as anything that requires abstract reasoning, or requires learning from very few examples.

This is why Phrasee is a perfect example of effective deep learning: it doesn’t rely on abstract or insignificant information. Instead, Phrasee leverages historical data and the results of automated experimentation, by applying deep learning networks to customer and market data in order to predict and generate top-performing language for your brand.

Take a trip inside the Phrasee brain to find out how deep learning can help you can say it better.

Image credit: Disney
via GIPHY

Want to learn more about AI? We’ve got everything you could ever want to know from A to Z – just click here.


Trying to get ahead of the competition?

Check out our Learning Hub for everything you’ll need to know!