Artificial Intelligence Overview

Overview of Artificial Intelligence (AI)

• AI is a branch of science focused on enabling machines to solve complex problems in a human-like manner.
• It involves mimicking human intelligence characteristics and applying these as computer-friendly algorithms.
• The approach can vary in flexibility and efficiency, impacting how "artificial" the intelligent behavior seems.
• From an intelligence perspective, AI aims to make machines act "intelligently," similarly to human expectations.
• The inability to distinguish computer responses from human responses is known as the Turing test.
• Intelligence requires knowledge and expert problem-solving, which restricts the domain for relevant knowledge.
• From a business perspective, AI provides tools and methodologies to solve business problems.
• From a programming perspective, AI includes studying symbolic programming, problem-solving, and search.
• AI programming languages include LISP.
• LISP was developed in the 1950s, is an early language strongly associated with AI
• LISP processes heterogeneous lists and features run-time type checking, higher-order functions, automatic memory management, and an interactive environment.
• PROLOG was developed in the 1970s, is also strongly associated with AI, based on first-order logic, and has features for limiting the search space.
• Object-oriented languages are used for AI programming, it includes objects, messages, data bundling, inheritance, and methods for manipulating data.
• The computer is interrogated by a human via teletype; it passes if the human cannot tell if there is a computer or human at the other end.
• Artificial Intelligence is a new electronic machine that stores large amount of information and process it at very high speed
• It is the ability to solve problems
• AI is the science and engineering of making intelligent machines, especially intelligent computer programs, related to understanding human intelligence.

Importance of AI

• AI is important in Game Playing: computers can play master-level chess using brute force computation.
• AI is important in Speech Recognition: speech recognition has reached a practical level, for example, United Airlines uses speech recognition for flight information.
• AI is important in Understanding Natural Language: the computer needs an understanding of the domain the text is about.
• AI is important in Computer Vision: requires partial three-dimensional information that is not just a set of two-dimensional views.
• AI is important in Expert Systems: expert systems embody knowledge from human experts to carry out tasks.
• One of the first expert systems was MYCIN in 1974, which diagnosed bacterial infections and suggested treatments, outperforming medical students/doctors if limitations were observed.
• Ontology included bacteria, symptoms, and treatments but excluded certain vital parameters.
• Heuristic Classification is a feasible expert system that categorizes information from various sources, such as advising on credit card purchases.

Applications of AI

• In Consumer Marketing, credit/ATM/store card usage feeds into AI algorithms for data analysis.
• Companies gather data weekly to search for patterns, identify customer segments, and track responses to new products.
• Biometric Identification uses cameras, fingerprint devices, and microphones.
• Biometrics includes face, eyes, fingerprints, and voice patterns.
• Biometric identification compares data from a person at a door with a stored library for identification.
• Learning algorithms improve matching processes through off-line database analysis.
• Intrusion Detection aims to secure computers using patterns of computer usage.
• The goal is to learn the "signature" of each authorized user in order to identify unauthorized users.
• Programs accomplish this by recording user commands and time intervals, modeling variability, and classifying new users based on similarity.
• Machine Translation addresses language problems in international business.
• Fully automated translation is challenging because it requires translating both the words and their meaning.

Overview of Machine Learning

• Machine Learning was first coined by Arthur Samuel in 1959.
• Machine learning is a subset of AI, that allows machines to learn automatically and improve from experience without explicit programming in order solve problems.
• Tom M. Mitchell defines Machine Learning as: "A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P if its performance at tasks in T, as measured by P, improves with experience E."

Machine Learning Definitions

• Algorithm: A set of rules and statistical techniques used to learn patterns from data and draw significant information.
• Model: The primary component of machine learning, trained by a machine learning algorithm.
• Predictor Variable: A feature of the data used to predict the output.
• Response Variable: The feature or output variable to be predicted using the predictor variable(s).
• Training Data: Data set used to build the machine learning model, helping identify key trends and patterns.
• Testing Data: Used to evaluate the accuracy of the trained model.

Machine Learning Process Steps

• Define Objective: Understand the required predictions.
• Data Gathering: Determine data needs, availability, and collection methods.
• Data Preparation: Address inconsistencies like missing values or duplicates.
• Exploratory Data Analysis (EDA): Understand patterns, trends, and correlations within the data.
• Building a Machine Learning Model: Split data into training and testing sets, and the model logic depends on the algorithm implemented.
• Model Evaluation & Optimization: Test a model using the testing data set.
• Predictions: final output that can be a categorical variable (True or False) or a Continuous Quantity.

Overview of Deep Learning

• Deep learning overcomes feature extraction challenges, it enables models to focus on relevant features autonomously.
• It mimics the brain's learning process using artificial neurons, inspired by the structure and function of the brain (artificial neural networks).

Deep Learning

• Deep learning automatically identifies important features for classification using deep neural networks.
• Deep learning works by fixating on local contrast patterns at the lowest level.
• Subsequent layers identify eyes, noses, and mouths.
• The top layer applies facial features templates.
• Deep neural networks compose increasingly complex features in successive layers.
• Deep networks can overcome Machine Learning issues through inferences from data sets lacking proper labeling.

Applications of Deep Learning

• In Speech Recognition: Siri, Apple's voice-controlled intelligent assistant which uses deep learning to improve services.
• Deep learning enables accurate acoustic models that can learn and adapt, enhancing assistance through better predictions.
• Google Translate uses Machine Translation to provide instant translation between more than one hundred human languages.
• Instant Visual Translation identifies image letters, translates them, and recreates the image with translated text.
• Automated Self Driven Cars: Google's WAYMO uses deep learning to program systems that learn from sensor data, enhancing perception and control tasks.

Types of Machine Learning Systems

• Machine Learning Types can be achieved by one of three approaches.
• approaches include: Supervised Learning, Unsupervised Learning, and Reinforcement Learning
• Supervised learning uses labeled data to train the machine, similar to a teacher guiding a student.
• Training data is labeled, teaching the machine patterns within dataset.
• Unsupervised learning trains using unlabeled data, allowing the model to act without guidance, figuring out patterns and differences.
• Reinforcement learning involves placing an agent in an environment to learn behavior through actions and rewards.

Type of Problems in Machine Learning

• Regression: the output is a continuous quantity
• Classification: the output is a categorical value.
• Clustering: assigns the input into 2 or more other clusters based on feature similarity.

challenges Faced By Machine Learning

• Analyzing data for building or training models and in 2024 the global machine learning market is expected to have grown by 43%
• Machine learning experts tackle challenges to apply ML skills and create applications from scratch.
• Poor Quality of Data: Absence of clean and unpolluted data affects learning processes.
• Under fitting of Training Data: Data unable to establish a precise relationship between input and output variables.
• Over fitting of Training Data: Machine learning model is trained with an massive amount of data.
• Machine Learning is a Complex Process: Includes continuous changes, analysing the data, removing bias etc.
• Lack of Training Data: Less amount of trained data results in biased predictions.
• Slow Implementation: Models takes a tremendous amount of time. Slow programs, data overload, and excessive requirements usually take a lot of time to provide accurate results.
• Irrelevant Features: Training data must contain relevant features, that involves feature engineering.
• Imperfections in the Algorithm When Data Grows: best model of the present may become inaccurate in the coming Future and require further rearrangement.