A field developing systems for tasks such as recognising images, predicting outcomes and generating text. AI does not automatically imply human understanding. A model's confident answer can still be wrong.
A field developing systems for recognition, prediction and content generation, among other tasks. It includes more than chatbots. This is the expanded name of AI; task performance does not establish consciousness or infallibility.
A historical name associated with Microsoft's Bing Chat. In November 2023 it became Microsoft Copilot. It was not simply a Turing NLG product; capabilities depend on the service's current version.
An OpenAI service for conversation and work with AI models, including text and information tasks. Tools vary by version and plan. Results need checking; ChatGPT is not the name of one fixed GPT-3 model.
An OpenAI model family developed to generate images from descriptions. It is not a generic name for every image generator. Check documentation for version availability; OpenAI's offering also includes GPT Image models.
The name of Google's voice assistant for questions and device control. Google is moving parts of the experience to Gemini; feature availability depends on the device.
The historical name of Google's chatbot, replaced by Gemini in February 2024. Bard used different models during its development. It should not be described as a separate current service permanently based on LaMDA.
The name of Google's AI model family and services using those models. The Gemini app succeeded Bard. Models and apps are different things; tools vary by product, version and plan.
An AI assistant and model family developed by xAI. Tools can support research and content generation, depending on availability. An advertised conversational style is not evidence of accuracy; outputs still need verification.
Tools and activities for processing, storing and communicating information. Business IT includes applications, devices, networks and their maintenance. It extends beyond programming.
A model trained on large datasets to process and generate language. Fluent output guarantees neither factual correctness nor access to current information.
A platform using generative models to create and edit visual material. Features, models and licences depend on plans and service terms. Generated output still needs quality and usage-rights checks.
A proposed file pointing LLM-based tools to useful material. It is not a universally adopted standard or a replacement for robots.txt and access control.
An AI service for generating and transforming images using instructions and reference material. Features and rights to use results depend on the version, plan and service terms.
Practices for maintaining machine-learning models, data, deployments and monitoring. They address changes in data and model quality after production launch.
A field of computational language processing, including classification and information extraction. Not every NLP method requires a large language model.
The historical name of Opera's browser AI assistant. Documentation now describes Opera AI using multiple models and tools. Features vary by version; Aria should not be assigned one permanent GPT-4 model or unlimited music and video generation.
Input and instructions supplied to an AI model. Context and response criteria affect output, but a good prompt guarantees neither correctness nor safety.