Artificial Intelligence: A Brief Overview
Artificial intelligence (AI) is fundamentally the replication of human
intellect in robots trained to think like people and copy their
activities [6]. The phrase also refers to any machine that has
characteristics similar to the human mind, such as learning and
problem-solving. AI’s strength lies in its ability to analyze,
comprehend, and learn from massive volumes of data, find patterns, make
predictions, and adapt to new information. At the heart of many AI
systems is the backpropagation algorithm, a mathematical technique used
to train artificial neural networks [7]. AI is a broad field with
many sub-areas, but at a high level, it can be divided into two
categories: narrow AI, which is designed to perform a specific task,
such as voice recognition, and general AI, which can understand, learn,
and apply knowledge across a wide range of tasks, much like a human. The
notion of AI has its origins in antiquity, with myths, legends, and
theories about artificial entities endowed with intellect or
consciousness by skilled artisans. However, as a scientific subject, AI
is relatively new. It was initially presented as a topic of study at the
Dartmouth Conference in 1956 when the term ’Artificial Intelligence’ was
created [8]. This signalled the beginning of AI as an autonomous
discipline. The subsequent decades saw a series of ups and downs, with
periods of intense excitement and funding (known as AI summers) followed
by periods of disappointment and reduced funding (known as AI winters)
[9]. Early AI research concentrated on rule-based systems and
symbolic thinking. However, when dealing with complicated, real-world
challenges, these techniques proved to be limiting. The development of
machine learning in the 1980s and 1990s transformed the area, allowing
computers to learn from data rather than writing explicit rules. This
revolution culminated in the creation of deep learning in the 2000s and
2010s, a technique based on the architecture of the human brain that
analyzes and learns from data using artificial neural networks [11].
The backpropagation algorithm, central to training neural networks,
involves the iterative application of the chain rule from calculus to
update the weights in the network [12]. It can be summarized by the
following equation:
\begin{equation}
\Delta W\text{ij}=-\eta\frac{\partial E}{\partial w\text{ij}}\nonumber \\
\end{equation}Where: Δwij is the change in weight connecting neuron ito neuron j , η is the learning rate, ∂E/∂wij is the partial derivative of the error function Ewith respect to the weight wij .
Deep learning, powered by backpropagation, has been responsible for many
of AI’s most exciting recent advancements, including self-driving
vehicles, voice assistants, and recommendation systems. Today, AI
technologies have permeated nearly every part of our lives and are
evolving at an unprecedented rate. As we approach a new age in AI
research and application, it is critical that we examine and exploit its
ability to interpret complex patterns in human behaviour, leveraging
tools such as the backpropagation algorithm to advance our understanding
and capabilities.