AI basics

Algorithm: A defined set of instructions that guides a machine in learning how to perform a specific task.
Artificial Intelligence: The overarching concept of machines emulating human intelligence through various features like human-like communication and decision-making.
Autonomous: When a machine can operate without requiring human intervention to perform its designated tasks.
Backward Chaining: A methodology where the model begins with the desired output and retraces its steps in reverse to uncover supporting data.
Bias: Assumptions made by a model that simplify the learning process for its assigned task. Reduced bias is usually preferred in supervised machine learning as excessive assumptions can adversely impact results.
Big Data: Refers to datasets that are excessively large or complex for traditional data processing applications to handle.
Bounding Box: Typically used in image or video tagging, it represents an imaginary box drawn around visual content, with labeled contents to aid the model in recognizing distinct object types.
Chatbot: A program designed to interact with people via text or voice, simulating human-to-human conversation.
Cognitive Computing: An alternative term for artificial intelligence, sometimes used to alleviate the science fiction connotations associated with AI.
Computational Learning Theory: A branch of artificial intelligence primarily focused on creating and analyzing machine learning algorithms.
Corpus: A large collection of written or spoken material used to train machines for linguistic tasks.
Data Mining: The process of analyzing datasets to uncover new patterns that can enhance models.
Data Science: An interdisciplinary field combining statistics, computer science, and information science to address data-related problems using various scientific methods, processes, and systems.
Dataset: An assembly of related data points, typically organized with uniform sequencing and categorization.
Deep Learning: A facet of artificial intelligence that emulates human brain functionality by learning from data structure rather than relying on programmed algorithms.
Entity Annotation: The act of labeling unstructured sentences with information to enable machine comprehension, often encompassing the identification of people, organizations, and locations in a document.
Entity Extraction: A comprehensive term for structuring data to make it machine-readable, which may involve human or machine-driven labeling.
Forward Chaining: A methodology where a machine starts with a problem and seeks potential solutions by analyzing various hypotheses relevant to the problem.
General AI: An AI capable of performing any intellectual task achievable by a human being, also known as strong AI.
Hyperparameter: Values affecting how a model learns, often set manually outside the model, sometimes used interchangeably with parameter.
Intent: A type of label found in training data for chatbots and natural language processing tasks, defining the purpose or goal of the communication.
Label: A component of training data identifying the desired output for a specific piece of data.
Linguistic Annotation: The tagging of a dataset’s sentences with their subjects in preparation for analysis, often employed in tasks such as sentiment analysis and natural language processing.
Machine Intelligence: A broad term encompassing various learning algorithms, including machine learning and deep learning.
Machine Learning: A subset of AI that concentrates on developing algorithms enabling machines to learn and adapt to new data without human intervention.
Machine Translation: The algorithmic translation of text, independent of human involvement.
Model: A general term referring to the result of AI training, produced by executing a machine learning algorithm on training data.
Neural Network: A computer system designed to function like the human brain, with existing neural networks capable of various tasks involving speech, vision, and strategy.
Natural Language Generation (NLG): The process through which a machine converts structured data into human-readable text or speech, typically occurring as the final step in communication.
Natural Language Processing (NLP): The overarching term for a machine’s capacity to perform conversational tasks, encompassing understanding, interpretation, and coherent responses.
Natural Language Understanding (NLU): A subset of natural language processing focused on enabling machines to discern the nuanced meanings in language, accounting for subtle nuances and grammatical errors.
Overfitting: A common issue in machine learning where the algorithm can only work on specific examples present in the training data, making it unable to generalize to new examples.
Parameter: An internal variable within the model that aids in making predictions, with values typically estimated from data and not manually set by the operator.
Pattern Recognition: The identification of trends and patterns in data, often intertwined with machine learning.
Predictive Analytics: Combining data mining and machine learning to forecast future events based on historical data and trends.
Python: A widely used programming language in general programming.
Reinforcement Learning: A training method that sets goals without specific metrics, encouraging the model to explore various scenarios for better results based on human feedback.
Semantic Annotation: Labeling different search queries or products to enhance the search engine’s relevance.
Sentiment Analysis: The process of identifying and categorizing opinions in text to determine the writer’s attitude toward a subject.
Strong AI: An area of research focused on developing AI with capabilities equivalent to the human mind. Strong AI is often used interchangeably with general AI.
Supervised Learning: A type of machine learning that uses structured datasets with inputs and labels to train and develop algorithms.
Test Data: Unlabeled data used to evaluate a machine learning model’s ability to perform its designated task.
Training Data: All data employed in the process of training a machine learning algorithm, specifically referring to the dataset used for training as opposed to testing.
Transfer Learning: A learning method involving teaching a machine a related task and then returning it to its initial work with enhanced accuracy.
Turing Test: A test, named after Alan Turing, assessing a machine’s ability to mimic human behavior and language. The machine passes if its output is indistinguishable from a human participant’s.
Unsupervised Learning: A training approach where the algorithm must make inferences from unlabeled datasets, helping it learn.
Validation Data: Structured like training data with inputs and labels, this data is used to test a newly trained model against fresh data, focusing on performance and identifying overfitting.
Variance: The extent to which a machine learning model’s intended function changes during training, with models having high variance prone to overfitting and reduced predictive accuracy due to heavy reliance on training data.
Variation: Alternate queries or utterances in conjunction with intents in natural language processing. Variation represents what individuals might say to achieve a specific purpose or goal.