OVERVIEW OF ARTIFICIAL INTELLIGENCE (AI)

Source: Deep Learning on Medium

OVERVIEW OF ARTIFICIAL INTELLIGENCE (AI)

This article was motivated by the presentation at the first AIsaturday Ilorin meetup organized by Adnan Adetunji (AdnanHaddy), the facilitator.

“My dream is to achieve AI for the common good”— Oren Etzioni

I would say Oren Etzioni read my mind when he said those words but I really did not know how I could achieve AI until I joined AISaturday Ilorin cohort at Malhub. Before then all I did was munch on AI and tech articles online, sometimes code for few minutes and take a break for a month or more before returning to it.

Intelligence is an attribute commonly associated with human being. Teaching or allowing computer to learn this attribute and actually mimic human in terms of intelligence can be referred to as artificial intelligence. However, the intelligence portrayed by machine/computer is artificial but machine can learn to take decisions on their own after taking several instructions and learning from experience.

The combo of words “Artificial intelligence” has been thrown around on the internet and trending for some time now. Artificial intelligence (AI) is being mistaken for machine learning (ML) and both are used interchangeably by some folks. Well, it is important to state clearly that machine learning can be better thought of as a subset of artificial intelligence. In order word, machine learning is simply a way of attaining AI.

AI from within

Machine Learning(ML) is achieved by training a model (a computer program) to learn as much from data in order to perform task pertaining to the provided data. ML can be classified into two viz: supervised learning and unsupervised learning. Supervised learning is achieved when data is being analyzed toward a specific target or getting a specific value. Depending on the target, Supervised learning can be divided into regression and classification task. Unsupervised learning involves identifying patterns in data without known targets. Reinforcement learning is the repeated steps in allowing agent/machine to decide the best option to select based on his current position. This form of learning has been used in practice for mostly game development.

Another subset of AI worth mentioning is Deep learning. Deep learning is associated with learning from enormous data but the method of learning is what differentiate it from machine learning. Hence, Deep learning is a subset of ML. Deep learning is achievable through Neural network. Neural network’s inspiration is from the human brain. It was designed to mimic the neurons in the brain by its arrangement of nodes and stacks of nodes called layers. The network is arranged in such a way that the output from one node will serve as input to another.

It is worthy of note to mention some fields. And they are:

1. Natural Language Process (NLP) which is the study of interaction between computers and human languages. The application of NLP encompasses Text classification and recognition, sentimental analysis and chatbot.

towardsdatascience.com

2. Computer Vision (CV): This is a field in computer science that is concerned with enabling computers see, identify and process images. Due to these automated process(es) on image(s) that computer must perform on its own, CV is therefore linked to AI.

Since images are seen as grid of numbers (between 0 to 255). These numbers are arranged in vectors and vector operations are perform. It is possible after training, for the computer to recognize images.

Everyday applications of AI are text to speech systems (google assistant in Android OS and Siri in iOS, Cortana in Windows OS), spam classifications in our emails, clustering and recommender system in ecommerce websites (and social media), Autonomous cars, facial recognition app embedded in our phones and so on. These applications could be disadvantageous to us to some extent.

Generative Adversial Network (GAN). GAN involves using a model to generate new examples that plausibly come from the same distribution of samples. GAN can be misleading in the sense that if one should engage in a search for image produced by a GAN algorithm, one can be in search of a ghost or a place that is not in existence at all. Sometime in December 2019, Facebook deleted accounts with suspected artificially generated profile pictures

Two profile pictures on Facebook suggested by Graphika to be an artificially generated pictures. The rectangles on the images are the marked areas of the images with weird inconsistency.

Two profile pictures on Facebook suggested by Graphika to be an artificially generated pictures. The rectangles on the images are the marked areas of the images with weird inconsistency.

If GAN can produce anime characters, short videos of unreal kids playing in a field, in the next 10 years, we foresee GAN been capable of producing a full movie of generated characters and actions not far from reality. The capabilities of GAN can be a turn on for most people who are interested in tech and want to have impact in the future through AI.

At this point, you might have a little trust issue with AI but

“AI is neither good nor evil. It’s a tool. It’s a technology for us to use.”— Oren Etzioni

Technology has never been good or evil, it’s people earnesting it that are either good or evil.

Contrary to the believe that knowledge is best gained with a basic-to-advance approach, some programming books start with instructions to code “Hello world” first, this is then followed by other steps of coding. These books advice that readers should code first without bothering about what the code does, that is learnt later. Human beings (especially in this century) are so used to their life of instant gratification. When a beginner is told to code first without knowing what the code means (just to see the output of the code), his curiosity of learning this language is suppressed. It is therefore advised to learn to output then learn the underlying working principles thereafter. Adnan Haddy advised that for self-study, one should learn linear algebra, calculus, statistics, take the Andrew Ng and CS231n Stanford course and read papers (to keep yourself abreast of the latest researches on AI). Also read AI books and blog posts. Lastly and most importantly, code, code and code, Join a AI group like a slack community (online) and offline like an AISaturday meetup in your community. I am glad to tell you that the advice actually is still working for me.

After learning so much about what AI is about and what we can achieve with it, I had prepared my mind to take this deep dive into the world of AI because I believe that with AI the achievable is endless.

“Artificial Intelligence will digitally disrupt all industries. Don’t be left behind” — Dave Waters