What is Artificial Intelligence?
Artificial Intelligence (AI) has emerged as a critical concept in today’s technology-driven world. Yet, its fundamental nature is not understood by many. This article aims to demystify the concept of AI in layperson’s terms, particularly for those outside the technical sphere.
Human intelligence is the ability to learn from experiences and reason with the knowledge to make informed decisions. We interpret our observations, analyse them, and deduce logical conclusions. This process is fundamental to how humans understand and interact with the world around them. AI strives to emulate certain facets of human intelligence in machines. It entails programming computers to analyse and learn from data, enabling them to reason, make decisions, and solve problems. Unlike human intelligence, which is rooted in organic experiences and cognitive functions, AI operates on algorithms and computational techniques. The goal of AI is not just to mimic human thought processes but to augment our capabilities, allowing us to tackle complex tasks more efficiently and effectively.
In today’s world, AI is often equated to machine learning, yet its scope extends far beyond that. AI is continually evolving and events like the AAAI Conference offer a glimpse into its expansive scope, showcasing various applications and theoretical approaches. My aim here is to formulate a more comprehensive and easily understood definition of AI for the general public.
Cassie Kozyrkov has eloquently described AI as a revolutionary change in how humans interact with computers. This characterisation is especially relevant, as it emphasises the role of AI in amplifying human capabilities through the power of computing. By utilising AI, we can execute tasks with enhanced efficiency, overcoming the limitations of human speed and precision. AI encompasses two primary components machine reasoning and machine learning, which I will elucidate further.
Machine Reasoning
The Travelling Salesman Problem (TSP) can exemplify machine reasoning in artificial intelligence. This scenario highlights the distinction between the ease of problem description and the complexity of problem-solving.
Problem Description: Imagine a situation where you must plan a journey through various cities in a country, visiting each city only once and aiming to find the shortest possible route. This is known as the travelling salesman problem. Describing this problem to another person is relatively simple. It can be done using everyday language, focusing on the cities to be visited and the distances between them. The real challenge lies in determining the most efficient route as the number of cities increases. While finding a route among ten cities might be manageable manually, the task becomes more complicated with 10,000 cities. In such cases, manually figuring out the optimal path is nearly impossible.
This is where machine reasoning becomes crucial. Modern computing languages allow us to state these problems succinctly to machines, which can then quickly solve these problems by applying logical and algorithmic thinking and determine an optimal or near-optimal solution. Although no algorithm is known that can guarantee finding an optimal solution for a larger size problem in a reasonable time, significant progress has been made in developing efficient methods to solve them quickly. This involves complex computations, heuristics, and decision-making that mirror human reasoning but are executed at a scale and speed far exceeding human capabilities.
In summary, machine reasoning in AI demonstrates the power of machines to tackle complex problems that are easy to describe but extremely challenging to solve, showcasing the advanced capabilities of AI in problem-solving and logical analysis.
Machine Learning
The second aspect of AI is learning, which can be illustrated by the challenge of describing someone’s appearance, such as an exotic animal, Quokka, a small marsupial native to Australia, to someone who has never seen one. You might mention its small size, rounded ears, and the fact that it resembles a miniature kangaroo. However, such a description is unlikely to fully capture the unique appearance and charm of the Quokka, including its facial expressions and distinctive movements. This difficulty in human-to-human description underscores the challenge of communicating such visual information to a machine.
In contrast to verbal descriptions, showing an image to another person usually results in immediate recognition, thanks to the complex processes in our biological brains. Similarly, we can teach machines to recognise and understand images by presenting them with numerous pictures. Through exposure to these images, machines employ pattern recognition and learning algorithms to develop an internal representation of what they “see”. This effectiveness of visual learning showcases the immense potential of AI in image recognition.
This approach underscores the core principle of machine learning: providing machines with data (in this case, images) and enabling them to learn and recognise patterns and details autonomously. Over time, these machine-learning models become adept at identifying objects in images. This capability has numerous practical applications.
Summary
As we delve into machine reasoning and machine learning, it becomes clear that AI is not just a technological advancement but a journey towards redefining problem-solving and knowledge acquisition. The synergy of these two aspects of AI — machine reasoning’s ability to solve complex problems and machine learning’s capacity to evolve from data — indicates a future where the collaboration between humans and machines creates unprecedented opportunities. Together, they are opening up exciting avenues to find solutions we have yet to think of.
