AI glossary: A simple guide to understanding the technology

29. March 2021 | Drooms Global


Artificial intelligence is everywhere. It is in our business software, the apps we use, the trains we take and the news we watch. While we often hear about it and even talk about it, different concepts surrounding AI are not always understood. So, what should you know about AI?

Artificial intelligence is constantly evolving. Its future will undoubtedly look very different to what we have come to know today. But for now, we provide a glossary of some of the most important and popular terms around artificial intelligence. These terms can help you understand artificial intelligence software and its applications.

 

Artificial intelligence

 

Although we know that artificial intelligence enables a machine to perform cognitive functions (like perceiving, learning, reasoning, and solving problems) interestingly, there is still no homogeneous definition for AI. Artificial intelligence is divided into three subfields:

• Artificial intelligence - the main category

• Machine learning - the first subcategory of AI

• Deep learning - the first subcategory of machine learning

Artificial intelligence has been used in the development of self-driving cars, chatbots and social media monitoring.

 

Artificial Narrow Intelligence

 

When articles talk about “weak AI”, they are referring to Artificial Narrow Intelligence (ANI). The term ‘weak’ is an inaccurate way to describe ANI however. ‘Weak AI’ lead to highly advanced solutions, where the idea is to augment or replace humans for specific tasks. For example, self-driving vehicles use Artificial Narrow Intelligence.

 

Artificial General Intelligence

 

The world is also working towards Artificial General Intelligence (AGI) or “strong AI”. The concept is highly debated, as some experts do not believe it is possible to achieve AGI. The end goal of AGI would be to achieve the same level of functionality as a human brain or to even surpass human intelligence.

 

Algorithm

 

Algorithms are often coded by humans and based on a set of instructions. For example, an algorithm can help make sense of datasets. Suggestions you receive on shopping platforms are based on algorithms coded to analyse data.

While most algorithms in the past have been coded by humans, AI can help machines create algorithms without human input. These algorithms tend to learn from the data they receive.

Algorithms use techniques like:

Classification which assigns categories to data points

Clustering which groups similar data into larger categories

Attributing which involves assigning certain characteristics to an object

Detection which classifies and localises one or multiple objects in a picture or text

 

Big data

 

The term itself refers to large quantities of structured and unstructured data that cannot be handled by standard data-processing software. As mentioned, data can be:

Structured: where it is clearly defined and has easily searchable patterns

Unstructured: where no identifiable patterns are easily found

 

Deep Learning

 

Deep learning is a flexible set of rules and techniques that allows the neural network to teach itself. Famously, image-recognition neural network Watson, could teach itself classes of images it was not trained to identify thanks to deep learning.

 

Machine learning

 

Artificial intelligence and machine learning go hand in hand today. Programs use machine learning to teach and alter algorithms to work better.

Learning can be:

Supervised: where algorithms run through training data to allow them to correct their performance. Usually, it means giving them the ‘right’ answers to speed up the process

Unsupervised: where algorithms analyse and find structures in the dataset without training or correcting

Reinforcement learning is a third type of learning. The objective here is to improve AI capabilities through feedback. The system will run different scenarios and human input will evaluate how well the system is doing. The AI program will then take on the feedback and adjust its behaviour.

 

Natural Language Generation (NLG)

 

This is a process where AI technology takes structured data and turns it into text. Typically, NLG systems are taught the different relationships between data points.

 

Natural Language Processing (NLP)

 

Natural Language Processing involves machines interpreting what human language means. It can also be used to generate natural language. NLP technology, a great example of the beneficial impact of artificial intelligence in business, in used in Drooms’ own product offering.

 

Neural network

 

As we work towards improving AI, neural networks become a key focus area. The aim is to create networks that are similar to the human nervous system and brain. Neural networking currently uses different stages of learning to allow AI to solve complex problems. In the first stage, the artificial intelligent model might simply worry about pixels on a screen. After the initial stage, the neural network passes the information further and AI tries to make sense of more data. The result would be deep learning.

 

Robotics

 

Applications of artificial intelligence often involve robotics. Artificial intelligence technology can be used in robotic devices to perform certain tasks. Robotics is defined broadly as a way to create and power physical automation.