1.4 AI and Related Fields
Artificial Intelligence (AI) is a field of Computer Science that seeks to automate intelligent behavior. This field is built on principles such as data structures, knowledge representation, algorithms, language, and programming techniques. Through AI, machines can simulate human intelligence and learn from data to make decisions or perform tasks. The potential applications of AI are vast and range from improving healthcare and transportation to enhancing customer service experiences.
1.4.1 Common Techniques Used in AI
Neural Networks: Neural networks are structures that can recognize patterns in inputs by training weights of “neurons”. They use supervised learning to build a function representation that approximates the input-to-output mapping.
Function approximation involves constructing an approximate function f’ that approximates f(x) given y1 = f(x1), y2 = f(x2), …, yn = f(xn). This function is typically smooth, with f’(x’) close to f’(x) for x’ close to x. Function approximation serves two purposes:
- Size: The expression of the inaccurate function can be extremely smaller than the right function.
- Generalization: The approximate function can be used on inputs where the function value is unknown.
In pathfinding, we use the function f(start, goal) = path to find the best path from start to goal. Rather than computing the paths using A* algorithm or other methods, we can use the function f to find the best path directly.
There is no point in generalizing f since we already know it completely. Reducing the representation size of f is the only potential benefit, but the algorithm is already simple and takes up little space. Additionally, neural networks produce a fixed-size output while paths vary in size.
Genetic Algorithms: Genetic Algorithms are used to optimize functions by exploring a parameter space to find solutions that score well based on a “fitness function”. This is an unsupervised learning problem where the goal is to find the value of x that maximizes or minimizes the function g(x), where x is a vector of parameter values.The process of path finding involves finding the path between a starting position and a goal. In optimization, g(x) represents the cost of the path and x represents the path itself. Hill-climbing optimization will only increase the cost of the path, and can lead to getting stuck at a “local maximum.” Genetic algorithms offer a better approach by maintaining multiple paths and using mutation and crossover. However, for path finding, algorithm A* can already find the best path, making optimization approaches unnecessary.
Genetic Programming involves breeding path finding algorithms, using fitness functions to rate their performance, instead of evolving new ones. .
Genetic algorithms may solve path-finding problems similar to neural networks.
Reinforcement Learning: Reinforcement Learning is a highly effective type of unsupervised learning that enables agents to learn from their experiences and share these insights with others. This innovative approach is similar to the renowned A* algorithm, as it evaluates the choices and actions that lead up to a specific state based on the reward or punishment that is associated with them. By leveraging the power of reinforcement learning, individuals and organizations can gain valuable insights and make more informed decisions that drive success and growth. The propagation is accomplished through a value function, similar to the heuristic function in A*. However, the value function is updated as the agents try new things and learn what works. Reinforcement learning and genetic algorithms have an advantage over simpler approaches as they balance exploring new things and exploiting learned information. Genetic algorithms use exploration via mutation, and reinforcement learning uses explicitly allowing the probability of choosing new actions. When it comes to solving the path-finding issue, it’s important to keep in mind that reinforcement learning is not the best approach. Instead, it’s better to use it as a tool for guiding agencies on how to navigate the game world effectively.
1.4.2 Related Fields of AI
The following are possible fields related to AI.
- Robotics
- Computer Vision (c) Image Processing
- Voice Recognition
- Neural Networks
- Artificial Intelligence – Robotics: Robots have become an essential part of various industries, from domestic to military, due to their practical application of rapidly advancing artificial intelligence. Unlike the assembly line robots of the past, today’s field and service robots are capable of operating in unstructured environments such as shops, homes, schools, and battlefields. Join us in developing flexible, autonomous software for robots. We need new machines and powerful algorithms to solve challenging problems. Students with skills in electronics, mechanics or software development and a passion for robots are welcome.
- Artificial Intelligence – Vision: Computer vision is a field that aims to duplicate human vision by electronically acquiring, processing, analyzing, and understanding high-dimensional data from the real world to produce numerical or symbolic information. This is performed by releasing symbolic information from image data using models made with the aid of geometry, physics, statistics, and learning theory. It also involves automating and integrating a wide range of processes and representations for vision perception.
Applications range from industrial machine vision systems, inspecting bottles on a production line, to AI research and robots that understand the world. The computer vision and machine vision fields overlap. Computer vision automates image analysis while machine vision combines it with other methods for inspection and guidance in industry.
Computer vision is a scientific discipline that studies the theory of artificial sike genetic algorithms, reinforcement learning should not be used for the path-finding issue itself, but instead as a direction for teaching agenystems extracting information from various forms of image data, such as video sequences, views from multiple cameras, or multi-dimensional data from medical scanners.
Computer vision is a technological specialization that applies approaches and models to make computer vision systems. Computer vision systems are used in additional applications such as:
Controlling processes, e.g., an industrial robot;
Navigation, e.g., Operated by a self-driving vehicle or a robotic mobile device;
Detecting events, e.g., This technology is suitable for tasks such as visual surveillance or counting the number of people present.
Organizing information, e.g., This is used for the purpose of indexing databases that contain images and sequences of images.
Modeling objects or conditions, e.g., medical picture examination or topographical modelling;
Interaction, e.g., As the input for a device used for interacting with computers and humans, and
Automated inspection is useful in manufacturing and other applications.
- Artificial Intelligence – Image Processing: We can use artificial intelligence techniques to process digital images including image fundamentals, enhancement, restoration, segmentation, edge detection, object recognition, representation, description, color processing, wavelets, multi-resolution processing, and compression.
- Artificial Intelligence – Voice Recognition: Voice recognition technology has been around since the 1950s. Since then, many companies, including Dragon Dictation, have experimented with it. The telecom industry has also created voice portals, which were meant to replace customer service reps, but ended up irritating cell phone users.
Despite technological advances, conversing with machines remained a sci-fi fantasy.
The automotive department has found a niche for the latest voice recognition software. Ford introduced the SYNC feature in 2007, which has become popular across its product line. Chevrolet advertised On Star’s ability to read Facebook feeds aloud during last year’s Super Bowl.
The field of speech recognition AI delves into the intricacies of the human thought process, seeking to replicate it in machines. This results in machines that are not only more intelligent but also more cost-efficient than natural intelligence. Natural Language Processing (NLP), on the other hand, harnesses the power of AI to enable seamless communication between humans and computers through natural language. The ultimate goal of NLP is to accurately comprehend the input provided and trigger specified actions by matching the input words with relevant keywords stored in its database.
- Artificial Intelligence – Neural Network: A neural network is a computer-based system that attempts to replicate the functions of the human brain by extracting specific properties of biological neurons and applying them to simulations. Although it is not yet possible to simulate the complexity of an actual brain, the components of an artificial neural network are designed to replicate the computing potential of the brain. The majority of artificial neural nets are structured according to the same fundamental structure, which includes input nodes that receive data, one or more layers of ‘hidden’ nodes that process the data, and output nodes that transmit information and provide activation values. This structure allows neural networks to learn from input data and generate output based on that learning, making them a powerful tool for a wide range of applications.
