Termos básicos que você precisa saber sobre Inteligência Artificial

Pyr Marcondes
10 de abril de 2019 - 6h37

Machine learning (ML): The branch of AI computing that involves training algorithms to perform tasks by learning from previous data and examples rather than explicit commands programmed by humans. Marketers often use ML when they want to optimize processes. Over time, these algorithms develop abilities and improve their own performance. Most AI technologies—including computer vision and natural language processing—are rooted in machine learning and its more complex descendant, deep learning. When companies talk about AI capabilities in their products and services, they are frequently talking about machine learning.

Deep learning: A complex branch of ML that involves building and training neural networks with multiple layers. Each network layer can use output from the layer above it to learn and make intelligent decisions on its own. Deep networks shine when sorting and classifying data and identifying anomalies in data patterns and excel at image and speech recognition, but they need more powerful machines than ML, and it’s often difficult for humans to understand how they work.

Neural networks: ML algorithms and computational models designed to function like neurons in the human brain. Neural networks are trained with specific sets of data points, which they use to guess at an answer to a query. The network’s guess is then compared with the correct answer for each data point. If errors occur, the “neurons” are tweaked and the process repeats itself until the error levels decrease. This algorithmic approach, called backpropagation, is similar to statistical regression.

Computer vision: Also called machine vision. The branch of AI that deals with interpreting and extracting meaning from visual elements in the real world, including printed characters or images such as faces, objects and scenes. It also incorporates image processing, pattern recognition and image understanding (turning images into descriptions that can be used in other applications). Computer vision underpins many up-and-coming innovations, including self-driving cars and cashierless stores.

Natural language processing (NLP): A branch of AI that deals with a machine’s ability to understand spoken or printed words in human (natural) languages, as opposed to computer programming languages. These technologies power conversational interfaces, including chatbots and virtual digital assistants and are heavily used by search engines for spam filtering and their ability to extract information from large and complex documents.

Natural language generation (NLG): A subset of natural language processing in which a computer makes decisions about how to comprehend a specific concept and put it into words. The technology can be used to automate manual processes for data analysis, such as personalized form letters and other types of communication at scale. It can also dynamically create communications— including basic news articles and real estate listings—that meet specific goals.

Chatbot: A computer program that uses a set of rules to conduct a speech- or text-based conversation with a human over an online chat interface. Chatbots are increasingly powered by AI and use NLP and NLG to mimic human conversation. Marketers often choose applications powered by this type of conversational AI when they want to interact with an audience.

Virtual digital assistants: A more sophisticated version of a chatbot, also known as an intelligent agent, voice assistant, virtual intelligent assistant, automated assistant or virtual agent. Such assistants can organize, store and output information based on the user’s location and can answer voice- or text-based queries from the user with information from a multitude of online sources (e.g., weather forecasts, maps, stock prices or transportation schedules). Examples include Apple’s Siri, Google Assistant, Amazon Alexa and Microsoft’s Cortana.

Recommender systems: Also known as recommendation engines. AI-driven information filtering systems that can automatically predict user preferences and responses to queries based on past behavior, one user’s relationship to other users, similarity among items being compared and context. High-profile examples of recommender systems include Amazon’s “frequently bought together” feature and Netflix’s CineMatch algorithm. Similar algorithms are also used by social networks such as Facebook, LinkedIn and Ancestry.com to find connections among people and data and to identify targets for marketing campaigns.

Predictive analytics: Programs that use a combination of techniques from data science, statistics and AI to analyze sets of structured and unstructured data, uncover patterns and relationships, and use them to make predictions about probable future outcomes and events. Predictive analytics models are closely related to prescriptive analytics models, which incorporate a predictive model but go a step further to produce