A branch of computer science that aims to create machines that can perform tasks requiring human-like intelligence. These tasks include problem-solving, pattern recognition, planning, and decision-making, among others.
A mathematical function applied to a neuron's output in a neural network. Common examples include the sigmoid, tanh, and ReLU functions.
A mechanism in deep learning models, particularly in NLP, that allows the model to focus on specific parts of the input when producing an output. Crucial in models like the Transformer.
An algorithm used in training feedforward neural networks for supervised learning. It calculates the gradient of the loss function concerning each weight by applying the chain rule.
Probabilistic graphical models that represent a set of variables and their conditional dependencies via a directed acyclic graph (DAG). They're used for reasoning under uncertainty.
A type of deep learning model that aims to improve the shortcomings of convolutional neural networks (CNNs), particularly in terms of handling spatial hierarchies between features. Proposed by Geoffrey Hinton.
A phenomenon in neural networks, especially in deep learning, where the network rapidly forgets previously learned information upon learning new data.
A thought experiment proposed by philosopher John Searle to challenge the notion that a computer program can possess genuine understanding or consciousness, even if it behaves as if it does.
An advanced search mechanism that goes beyond literal keyword matching to understand the context or concept behind a search query. It may use semantics, related terms, and underlying ideas to produce relevant results.
An AI approach emphasizing the importance of neural networks and parallel processing. It's based on the idea that cognitive processes are the emergent properties of interconnected networks of simple units.
A subset of machine learning that uses multi-layered neural networks to analyze various factors of data. It's particularly effective in tasks like image and speech recognition.
In NLP, embeddings are dense vector representations of words or phrases. These vectors capture semantic meaning, and words with similar meanings tend to have close vector representations.
An area of research aiming to make the decision-making process of AI systems clear and understandable to humans. It contrasts with "black-box" models where the decision process is not easily interpretable.
Optimization algorithms based on the process of natural selection. They are used to find approximate solutions to optimization and search problems, evolving solutions over time.
A theory and goal in AI research aiming for a harmonious understanding between humans and machines. It emphasizes the importance of machines not just processing information but also relating to human contexts and emotions.
A search mechanism that identifies content based on specific words or phrases. It contrasts with more advanced search mechanisms like conceptual search.
A type of machine learning model designed to understand, generate, or translate human language. Examples include OpenAI's GPT series.
A subset of AI where computers are trained to perform tasks by learning patterns from data rather than being explicitly programmed.
A controversial hypothesis proposed by Rupert Sheldrake that suggests a kind of collective memory in nature, which could influence the structures of systems and organisms over time. It's more metaphysical and hasn't been widely adopted in AI but has been discussed in relation to collective learning systems.
A subfield of AI that focuses on the interaction between computers and humans through natural language. The goal is to enable computers to understand, interpret, and generate human language in a valuable way.
A computational model inspired by the way biological neural networks in the human brain work. It's a fundamental building block in many deep learning models.
The ability of neural networks, both biological and artificial, to change their connections and behavior in response to new information, sensory experiences, or damage.
Proposed by Allen Newell and Herbert A. Simon, it states that a physical symbol system has the necessary and sufficient means for general intelligent action.
An area of machine learning where an agent learns to make decisions by taking actions in an environment to maximize a reward. It's inspired by behavioral psychology and has applications in areas like game playing and robotics.
Proposed by Marvin Minsky, it's a theory that intelligence is not the product of any singular mechanism but arises from the interactions of a diverse range of simple mechanisms.
A paradigm that studies collective behaviors from the local interactions of decentralized and self-organized systems. Examples include the flocking behavior of birds and the behavior of ant colonies.
An approach to AI that focuses on symbol manipulation and rule-based logic to solve problems, as opposed to the statistical methods seen in modern machine learning.
The process of converting a sequence of text into individual tokens (usually words or subwords). It's a common first step in NLP tasks.
A machine learning technique where a pre-trained model is fine-tuned on a new, similar task. This allows for leveraging knowledge from one task to improve performance on another.
A measure of a machine's ability to exhibit intelligent behavior indistinguishable from that of a human. Proposed by Alan Turing in 1950.
In the context of AI, vectors often refer to numerical representations of data. In NLP, word embeddings or sentence embeddings are often represented as vectors in a high-dimensional space. The spatial relation of these vectors can reflect semantic meaning.
Techniques in machine learning where models are designed to perform tasks without any examples (zero-shot), with only one example (one-shot), or with very few examples (few-shot).
