Demystifying Artificial Intelligence - What is Artificial Intelligence?

In the second part of our series, we look at what Artificial Intelligence is and how it has evolved.

Artificial intelligence (AI) refers to the development of computer systems or machines that can perform tasks typically requiring human intelligence. This includes problem-solving, learning, reasoning, pattern recognition, and understanding natural language. In this section, we will explore the definition and history of AI, its evolution from rule-based systems to machine learning, the various AI disciplines, and the impact of AI on different industries.

Image Generation Then and Now

The image on the left was created several months ago. The image on the right was generated in March of 2024 using the same prompt.

Definition and History of AI

While stories of Artificial Intelligence are as old as science fiction, the practical exploration of the field began in the mid-20th century with the work of pioneers such as Alan Turing, John McCarthy, Marvin Minsky, and others. In 1950, Turing’s famous “Turing Test” set the stage for evaluating a machine’s ability to exhibit intelligent behavior indistinguishable from that of a human. Over the years, the field has progressed through various phases, characterised by optimism, skepticism, and resurgence, driven by advancements in technology and algorithmic breakthroughs.

Before we throw in the towel and cede absolute control to our new robot overlords, it is important to understand two distinctions when we discuss AI:

Artificial General Intelligence (AGI)

AGI, sometimes referred to as “strong AI,” represents the aspirational goal of creating a machine or system that possesses human-like intelligence, capable of understanding, learning, and applying knowledge across a wide range of tasks and domains. In essence, AGI would exhibit cognitive abilities similar to those of humans, enabling it to independently solve complex problems, adapt to new situations, and exhibit a general understanding of the world.

In contrast, Narrow AI, also known as “weak AI,” refers to AI systems designed for specific tasks or applications, often excelling in their designated area but lacking the broader cognitive abilities associated with AGI. Examples of Narrow AI include speech recognition systems, recommendation algorithms, and image recognition software, which are all highly proficient in their respective domains but unable to transfer their learned knowledge to unrelated tasks. Consequently, the primary distinction between AGI and Narrow AI lies in the scope of their intelligence and applicability, with AGI representing a versatile, human-like intelligence and Narrow AI focusing on specialised, task-specific capabilities.

Despite outward appearances, ChatGPT is not AGI. Though, a recent paper suggests it may contain elements of it (Sparks of Artificial General Intelligence: Early experiments with GPT-4).

The Evolution of AI: From Rule-Based Systems to Machine Learning

Early AI research focused on rule-based systems that relied on explicitly programmed rules and heuristics to perform tasks like playing chess or solving mathematical problems. While rule-based systems demonstrated some success in well-defined domains, their limitations became apparent when dealing with complex, real-world scenarios.

The emergence of machine learning in the 1980s and 1990s marked a significant shift in the AI landscape. Machine learning enabled computers to learn from data, identify patterns, and make predictions without explicitly being programmed to do so. This approach proved more effective at handling the complexities and uncertainties of real-world situations, fueling AI’s rapid advancement.

However, there was insufficient data available at the time to train the algorithms, and the available computers were not powerful enough to run them. AI systems, particularly machine learning models, rely heavily on large volumes of data to learn patterns and make predictions. In the 1980s and 1990s, the availability of digital data was limited, as the internet was still in its infancy and many sources of information were not yet digitised. This lack of data hindered the ability of AI algorithms to learn effectively and achieve high levels of accuracy.

Further, AI models - especially the more complex ones - require significant computational resources to process and analyse data. During the 1980s and 1990s, the available hardware (e.g. processors and memory) was not powerful enough to handle the complex calculations and large-scale data processing required by advanced AI algorithms. This limitation made it challenging to develop and deploy sophisticated AI systems efficiently.

Despite these challenges, researchers in the field of AI continued to explore new methods and techniques, leading to the development of more efficient algorithms and laying the groundwork for future progress. The eventual explosion of digital data, spurred by the growth of the internet, and the advancements in hardware technology, such as GPUs and specialised AI chips, provided the necessary resources for AI to flourish. These developments have allowed AI to make significant leaps in recent years, leading to breakthroughs in machine learning, deep learning, and the emergence of powerful tools like large language models.

AI Disciplines: Natural Language Processing, Computer Vision, Robotics, and More

AI is a multidisciplinary field encompassing various areas of study, including:

• Natural Language Processing (NLP): The development of algorithms and models to enable computers to understand, interpret, and generate human language.

• Computer Vision: The study of techniques that allow computers to analyse, process, and understand visual information such as images and videos.

• Robotics: The design, construction, and operation of machines capable of performing tasks autonomously or semi-autonomously, often by mimicking human or animal behavior.

• Expert Systems: AI applications designed to replicate human expertise in a specific domain, such as medical diagnosis, financial planning, or legal advice.

• Neural Networks and Deep Learning: AI techniques that leverage artificial neural networks (ANNs), inspired by the human brain’s structure, to learn and process data.

• Reinforcement Learning: A type of machine learning where an AI agent learns to make decisions by interacting with its environment and receiving feedback in the form of rewards or penalties.

AI Applications and Impact on Industry

AI’s transformative potential has been recognised across various industries, leading to numerous applications:

• Healthcare: AI-powered diagnostic tools, personalised medicine, drug discovery, and robotic surgery.

• Finance: Fraud detection, algorithmic trading, risk management, and customer service chatbots.

• Retail: Inventory management, personalised recommendations, and customer behavior analysis.

• Manufacturing: Quality control, predictive maintenance, and optimisation of production processes.

• Transportation: Self-driving cars, traffic management, and route optimisation for logistics.

• Entertainment: Content recommendation algorithms, game development, and virtual reality experiences.

As AI continues to evolve, its impact on various industries and society as a whole is expected to grow, presenting both opportunities and challenges that we must address responsibly