AI FOR DUMMIES

Ever wondered how Google maps can find the shortest route between two points? Ever wondered how your YouTube feed is always relevant to your interests? AI, otherwise known as artificial intelligence, is behind all our daily conveniences and is to thank. The fact of the matter is that AI is all around us. For those of you who do not know what AI is or how it works, read on to find out.

Think of AI as a large container, and in that container is machine learning. And within machine learning is deep learning. Each subset has its own purpose as illustrated by the diagram

Artificial Intelligence

Any time a machine/computer system mimics human behavior, we call it AI. AI has three different types and are classified by “strength.”

1. Narrow AI: when a machine has superior performance to a human when doing one specific task.

2. General AI: when a machine is similar in its performance to a human in any intellectual task.

3. Strong AI: when a machine has superior performance to a human in many tasks.

Machine Learning

Keep these words in the back of your head for now:

· Labelled data: data where we know the target answer and the data object is fully recognized

· Unlabeled data: data where objects are undefined and need to be manually recognized

In machine learning, the algorithms are trained using earlier experiences and other examples to make predictions or decisions.

There are 4 types of machine learning that differ from each other based on the data they take in and the way they are trained:

1. Supervised learning: The system needs both an input and an output for training. Labelled data is fed into the system and is checked with the corresponding output. Application: categorizing emails as relevant or spam

2. Unsupervised learning: The system isn’t explicitly trained to find the right answer but is instead trained to find hidden patterns in the data inputted. It takes in unlabeled data and makes it more readable and organized, which makes patterns and anomalies more obvious.

Application: marketing

3. Reinforcement learning: The system isn’t trained, rather it learns on the basis of reward and punishment.

Application: search engines

4. Semi supervised learning: The system is trained using both labelled and unlabeled data. Much of the data used is unlabeled. This is because labelled data is much n more expensive.

Application: web crawler (sorts webpages)

Deep learning

Deep learning uses artificial neural networks (ANN) that mimic the human brain, as shown on the right. An ANN is made up of 3 parts: an input layer, hidden layers, and an output layer. The connection between each layer is given a random weight. The hidden layer is where all the magic happens; it processes the data coming in from the input layer, finding complex patterns that would take too long for a human to find.

Deep learning models are trained in a specific way. Large amounts of unlabeled data are fed into the ANN, and the system then identifies the data. Once enough data has passed through the system, some labelled data will be used the test the accuracy of the outputs. If the outputs are not accurate enough, a process called back propagation takes place. Back propagation is when the weights of the connections are adjusted until the desired output is achieved.

Deep learning is used in many ways, ranging from facial recognition software to photograph enhancements and text mining to name a few.

I hope this article has helped your understanding about Ai and how it works.