Artificial Intelligence: What is it?

Computism, by Grace Magny-Fokam
6 min readAug 15, 2022

--

The term “Artificial Intelligence” is often thrown around in movies and TV shows to showcase a world that has been conquered and run by robots. But outside of the context of science fiction stories, do we really know what artificial intelligence is? In reality, it’s quite different from what is shown in dystopian stories, so I’m here to break it down in a way that’s easy to understand.

Introduction

Artificial Intelligence (AI) is a sub-field of computer science that involves creating intelligent machines that can mimic human behavior. It’s the science by which machines learn patterns from large volumes of data and use that knowledge to perform tasks. AI operations are currently used in a variety of fields, such as entertainment, healthcare, marketing and sales, business analytics, etc.

Applications of Artificial Intelligence

Some examples of the applications of AI include:

  1. Recommendation systems: For instance, when Netflix recommends a TV show or movie to you based on the things you’ve previously watched, it’s using an algorithm in order to give you those recommendations.
Netflix’s Recommendation System

2. Smart Assistants (like Siri and Alexa): When you ask your smart assistant a question, AI is used to interpret what you’ve said and provide an output or response to your inquiry.

Apple’s Siri

3. Self-Driving Cars: In recent years, you may have heard of innovations from big tech companies like Google and Tesla Motors in the manufacturing of self-driving, AI-controlled vehicles. Recent developments in AI technology prove that AIs can facilitate partial or fully automated driving.

Google Driverless car

4. Spell/Grammar checking systems: Programs like Grammarly use AIs to automatically correct grammar and spelling errors in word processors.

Grammarly spell checker

5. Targeted online ads: Web browsers and social media platforms track the topics you research and are interested in, and use that data in an AI to show you ads that the AI thinks you are more likely to click.

Ad targeting AIs on social networks

Categories of AI

Under the umbrella of AI, there are 2 categories of AI that are based on an AI’s capabilities: Narrow AI and Strong AI.

Narrow AI, also known as weak AI, is an AI system that is limited to one specific area of function and isn’t capable of performing tasks outside of those limitations. This form of AI is quite common and is easily used to automate tasks. An example of this type of AI is any smart assistant like Siri from Apple. Siri can answer the user’s questions and navigate throughout your device, as long as what you’re asking Siri to do is one of its pre-programmed capabilities. If you were to ask Siri to do something that’s outside of that scope, however, it wouldn’t be able to do it because it wasn’t designed for that purpose.

Strong AI machines have general intelligence and can apply that intelligence to any task that you throw at them. In that sense, they are capable of dealing with a larger variety of tasks and can complete them using pre-existing information that’s been stored. These machines are more akin to the ones shown in science fiction stories, as they are much better at mimicking human intelligence.

Types of AI

Aside from these categories, there are 4 types of AI that are classified based on their functions: Reactive Machines, Limited Memory AI, Theory of Mind AI, and Self-aware AI.

Reactive machines perform basic operations using an input to produce an output. In these cases, machines don’t have to learn anything. An example of this would be if your input was a selfie of yourself and the output was your face being identified and boxed in using computer vision. The model didn’t have to learn anything from past data.

Selfie example of reactive machines

Limited Memory systems learn from past data in order to make predictions. The memory of these systems is temporary, though, so the data used is volatile and doesn’t save long-term. An example of the application of a limited memory AI system is in Tesla Motors’ self-driving cars. The car’s AI uses data that is collected from the environment surrounding the car in order to predict the best course of action at that moment. Once the prediction has been made and executed, however, the data that was collected from the surrounding environment is discarded.

Tesla’s self-driving interface

Theory of Mind AI is the concept of how intelligent systems interact with human emotions, sentiments, and thoughts. A semi-popular example of this is Sophia, an AI-powered robot created by Hanson Robotics in Hong Kong. Sophia is a humanoid that’s capable of exhibiting human emotions and behaviors and has reached worldwide fame for its ability to talk and act in the ways that a human would.

Sophia, created in 2016 by Hanson Robotics

Self-Aware AIs are still hypothetical since a fully self-aware AI that can totally replicate the behavior of a human brain has yet to be invented. In the evolution of AI, this is the most advanced state that any AI system can be, since the overall goal of AI is to mimic the human brain and its behaviors.

What if AIs were to become self-aware? | Alltime10s on YouTube

Subfields of Artificial Intelligence

Machine Learning (ML): ML is a subset of AI that allows computers to automatically gather and learn information from experience and use that knowledge to perform tasks. There are a few steps to follow when implementing an ML model including data acquisition and pre-processing, choosing and training a model, evaluation, parameter tuning, and prediction.

Deep Learning: Deep learning is a subset of machine learning in which computing models use deep neural networks to self-learn in order to perform tasks. These networks have numerous layers that are used to break down input data and learn from them.

Example Deep Learning model of George Washington

Neural Networks: A neural network is a network of algorithms that identify relationships in a set of data. Neural networks mimic how the human brain works and can extract features from data using their hidden layers. The image above shows an example of a simple neural network and how information is extracted in the hidden layers before it is output in the output layer.

Natural Language Processing (NLP): NLP allows computers to interpret and process human speech and language. This language can be verbal or written. Going back to my previous example with smart assistants like Siri, they use NLP in order to decode, process, and respond to human speech.

Computer Vision: Computer Vision allows machines to use their skills in deep learning and pattern recognition in order to extract patterns from images or videos. A very notable use of computer vision and image processing is in MRI scans and X-rays, in order to detect anomalies in the human body.

This image shows computer vision being used for object detection.

--

--