Creating a marketing platform built around Artificial intelligence, we are asked about the distinction in these definitions. So, this article attempts to help you understand these definitions and how they can be applied to marketing.
I have also applied each definition to marketing to show a simple application of how it can be used and its benefits.
This is a field that changes by the day but you can also read some great articles. One I particularly like is by Nvidia who make the graphics cards that power a lot of modern AI. Even down to powering self-driving cars by companies like Audi, Uber, and VW.
Artificial Intelligence (AI)
The term Artificial Intelligence was coined in 1956 and refers to a machine that is programmed to mimic the intelligence of the human mind through doing similar tasks. Then you get the concept of General and Narrow AI.
General AI refers to Artificial Intelligence designed to do a lot of things together. An example of this is the AI that Metigy uses to make its recommendations or to derive insights from the data. They are “general” as they actually combine a lot of our Narrow AI concepts.
Narrow AI Focuses on doing one task well. An example of this is the Sentiment Analysis we run. Whilst not simple, it is one task that the machine is trained to do very well and thoroughly.
How the machine is ‘trained’ or ‘programmed’ is where it gets interesting and is covered in Deep Learning and Machine Learning.
This is a simpler term for machines learning to do something on their own using some basic rules and massive amounts of data to look for patterns. The rules give it some hints as to what to look for or focus on. The more data the machine looks at, the more accurate it becomes at identifying that thing.
An example of Machine Learning in Marketing is what Metigy does looking at why a post worked or not. Our machine learning algorithms trawl through the posts we collect looking for basic patterns such as of words, tone, sentiment, and posting times.
Another example is audience analysis and profiling of that user. For example are they known or unknown? Are they likely Male or Female? Age group, location, etc. All information that can be obtained by looking at enough data to a degree of accuracy.
These patterns are there to help us identify why a post worked and help us start to recommend improvements to our users to help boost their posts organically.
Applying machine learning in this way can yield some unexpected results. An example occurred when scientists were testing it on solving a circuit board problem. The AI found an interim gate state that no one knew existed but meant the machine solved the problem more efficiently than thought possible.
This is the most modern take on Artificial Intelligence training and is a newer version of Machine Learning. At its core, it’s attempting to take an approach more akin to how Neurons in the Human Brain work.
Where it’s not possible – yet- to completely mimic the human brain, this approach breaks down a problem into layers. Take the example of an image. In Deep Learning, this is broken up into pieces and then handed to the next layer with each layer giving a level of ‘trust’ to its answer. Until you get a final answer at the end with a degree of accuracy. The more often this is done the greater the accuracy.
It can also lead to unexpected and sometimes scary results. Google has spent a lot of time working on their AI. They have nailed identifying important things such as Cats in photos. Down to discovering the first solar system outside of our own with 8 planets. They also have a dream AI built on a neural net that creates fantastic – and sometimes haunting – images.
In Marketing, this same deep learning can be applied in a lot of ways.
But, a simple example used by Metigy is Topic Analysis. That is to say what is a post or comment talking about? The algorithm works by scoring words and at each layer starts to build an association. That association builds on previous understanding looking for the connections derived from other posts, the community and even the images used – all of which have been analysed using Narrow AI that is pre-trained.
We will cover more articles in depth on AI, but this article gives you a good introduction to the differences of Machine vs Deep Learning and what AI is.
There are obviously those with concerns about how this technology could replace us. But at Metigy, we believe it can be used to help people do their jobs even better. We do not want to replace people. We want to help them enjoy their jobs more with our marketing assistant.
- The Difference Between Artificial Intelligence, Machine Learning, and Deep Learning by Calum McClelland
- What Is The Difference Between Deep Learning, Machine Learning and AI? on Forbes
- What’s the Difference Between Artificial Intelligence, Machine Learning, and Deep Learning? by Nvidia
This post originally appeared on the Metigy Blog on 13th March 2018