Glosario de IA

Comparing two or more variations of a product, tooltip, interface, or model to choose the best based on data rather than feel.

The principle that clear people, teams, or organizations should be held accountable for the decisions, errors, and consequences of an AI system.

A quality metric that shows what percentage of predictions the model made correctly.

An approach to speech recognition that associates an audio signal with likely sounds, syllables, or words.

Mathematical functions in a neural network that help it find complex nonlinear dependencies.

A machine learning method in which the model itself selects the most useful examples for human labeling.

A class of reinforcement learning methods where one component selects actions and another evaluates their utility.

The physical components of a robot or system that translate an algorithm's command into an action: moving, pressing, turning, or grasping.

An optimization algorithm that automatically adjusts the neural network training step and reduces dependence on manual selection of the learning rate.

An adaptive optimization algorithm that adjusts the learning step separately for each model parameter.

One of the most popular neural network training algorithms that adaptively adjusts parameter updates.

A statistical metric that evaluates the quality of a regression model, adjusted for the number of features.

Ways to fool a model with specially selected input data that looks normal to a human but confuses the algorithm.

An approach in robotics in which the system learns to understand what actions are possible with objects and the environment.

A modeling method where the behavior of a system is described through many individual agents with their own rules.

AI systems that can plan actions, use tools, and complete a task in a few steps with limited human intervention.

Sets of libraries and rules that help create AI agents with tools, memory, roles and execution scripts.

A data grouping method that gradually groups similar objects into clusters.

Specialized computing devices that speed up the training and execution of artificial intelligence models.

A European regulation that introduces risk-based rules for the development, sale and use of AI systems.

AI-based programs that can perform tasks through planning, memory, tools, and sequencing.

A field of study and practice that attempts to make AI behavior safe, beneficial, and consistent with human intentions.

A situation where an AI system systematically produces more favorable or unfavorable results for specific groups, cases or traits.

A direction where AI is used for environmental, social and economic problems, and also evaluates its own impact on resources.

Software interfaces through which applications connect to AI models, services for generating, recognizing or analyzing data.

Creation of pictures, illustrations, concepts and visual materials using neural networks based on text, reference or settings.

AI-powered digital assistants that answer questions, help with tasks, texts, searches, code or work organization.

Reviewing an AI system for quality, risk, safety, fairness, compliance, and suitability for real-world use.

The process of regularly reviewing an AI system, its data, behavior, documentation, and consequences of use.

US policy framework on protecting people when using automated systems and artificial intelligence.

A debated concern that AI-related valuations, investment, and infrastructure spending may be overheating relative to current revenue, adoption, and durable business value.

A set of measures that limit the capabilities of an AI system so that it does not exceed safe behavior limits.

Expanding access to AI technologies for people, small businesses, developers, education and organizations without large resources.

Panels of experts or internal committees that evaluate risks, rules, and controversial decisions in the development and implementation of AI.

Sets of principles and rules that help develop and use AI safely, fairly and transparently.

A system for storing, reusing and managing data features that are used in machine learning.

Application of artificial intelligence to analyze, forecast and reduce climate risks.

Texts, images, audio, video, code or other materials created in whole or in part using artificial intelligence.

Structures, policies and processes that help an organization securely develop, implement and control AI.

Rules, filters and technical barriers that help keep the AI ​​system within safe and acceptable limits.

Errors in which the AI ​​confidently produces fictitious, inaccurate or unconfirmed information.

Principles for the safe, fair and responsible use of AI in medicine and related services.

The use of artificial intelligence in defense, intelligence, cyber and autonomous military systems.

Laws and regulations that govern the creation, sale, implementation and use of artificial intelligence systems.

Ability to understand the capabilities, limitations, risks and rules for the safe use of artificial intelligence.

Platforms where you can find, compare, buy or connect AI tools, models, agents and ready-made solutions.

Model control processes throughout the entire life cycle: from selection and training to launch, monitoring, updating and decommissioning.

A platform or directory where artificial intelligence models are published, searched, downloaded, tested or connected.

A system for recording models, their versions, statuses, metrics, owners and change history.

Numerical parameters of a neural network that determine how the model transforms input data into a response.

Creating melodies, arrangements, beats, background music or vocal fragments using neural networks.

Coordinating models, tools, data, and process steps so that the AI ​​system performs a complex task in a controlled manner.

A set of rules, goals, and decisions by an organization or government about how to develop, use, and control AI.

Government, industry, and corporate regulations that limit and guide the use of artificial intelligence.

An area that deals with reducing the risks from errors, abuse, unpredictable behavior, and harmful effects of AI systems.

Protecting AI systems from attacks, leaks, abuse, unauthorized access and dangerous behavior.

A hypothetical moment when the development of artificial intelligence becomes so fast and powerful that the consequences are difficult to predict.

A document in which an organization discloses information about the operation, risks, data, limitations and controls of AI systems.

Tags or signals that help determine that content was created or modified by artificial intelligence.

A period of decline in interest, funding and expectations for artificial intelligence after inflated promises and weak results.

Using AI to perform repetitive tasks, transfer data between services, and support multi-step work scenarios.

An access model in which artificial intelligence functions are provided through a cloud service, API, or ready-made web interface.

British mathematician and one of the key scientists whose ideas became the basis of computer science and thinking about artificial intelligence.

An approach in which the decisions of algorithms must be controllable, verifiable, and associated with responsible people or organizations.

Techniques that help reduce unfair or systematic biases in data, models, and automated decisions.

A situation where an automated system unfairly makes people worse off because of attributes, data, or decision rules.

The ability to explain how an algorithm works, what data it uses, and why it makes certain decisions.

Scientist, teacher and entrepreneur known for his contributions to machine learning, online education and the popularization of AI.

Methods of searching for unusual events, objects, or patterns that differ markedly from normal data behavior.

Search for unusual actions, events, or network activity that may indicate an attack, leak, or compromise.

An artificial intelligence company known for its Claude family of models and its focus on AI security, controllability, and reliability.

A data streaming platform that helps systems share events in real time.

Design and creation of software interfaces through which services exchange data and functions.

Addresses or API routes through which the application accesses a specific service function.

A method of finding frequently occurring relationships and rules in data sets, especially shopping carts and transactions.

A hypothetical AI that can solve a wide range of tasks at or above human level, rather than just a narrow, specialized task.

The general name for technologies that enable computers to perform tasks typically requiring human intelligence.

Machine learning models inspired by a network of interconnected nodes that learn to transform data into the desired output.

A hypothetical AI that is significantly superior to humans in almost all intellectual tasks.

A specialized chip designed to perform a specific type of calculation as efficiently as possible.

Machine learning methods that look for consistent relationships between elements, events, or features in data.

A reinforcement learning method where multiple agents learn and update a common model in parallel.

Neural network components that help the model identify important parts of the input data and take into account the relationships between them.

A classifier evaluation metric that shows how well the model separates positive and negative examples at different thresholds.

A task in which the AI ​​determines the type of sound, event, speech, music, noise, or other audio condition.

Methods for analyzing, enhancing, transforming and extracting information from audio data.

A technology that overlays digital objects, cues, or data onto an image of the real world.

Verifying the identity of a user, service or device before accessing the system.

Checking what data and actions are allowed to an already authenticated user or service.

Neural network models that learn to compress data into a compact representation and restore it back.

An area of ​​AI where systems attempt to draw logical conclusions, prove statements, and solve problems using formal rules.

Technology that turns spoken speech into text using audio processing and language models.

Transferring repetitive actions to a system that performs them according to rules, data or AI decisions without constant manual intervention.

An approach in which the system helps automatically select models, features, parameters and training stages.

The ability of a robot, drone, car or other system to independently navigate and move in space.

Models that predict the next element in a sequence based on previous elements.