AI Glossary
- Artificial intelligence (AI): A machine’s ability to perform cognitive functions, such as learning, problem solving, and reasoning.
- Chatbot: A software that simulates human conversation through text or voice interactions. Chatbots typically leverage natural language processing (NLP) to understand a user’s input and provide a human-like response.
- ChatGPT: A large-scale chatbot powered by OpenAI’s GPT 3.5 large language model (LLM). ChatGPT was launched in November 2022 and drove massive interest in Generative AI.
- Foundational model: Base models trained on vast amounts of data. Other applications are built on top of foundational models.
- Generative AI: Algorithm-powered technology that can produce different types of content, including text, video, images, and more. Generative AI is a subset of artificial intelligence, but because it is powered by LLMs
- Hallucination: The term commonly used to describe when a large language model (LLM) makes up incorrect information but present it as if it were a true fact. Hallucinations are caused by many reasons, including outdated or insufficient data, idioms, and training limitations.
- Large Language Models (LLMs): AI models trained on massive data sets (billion or trillions of inputs), enabling them to learn language patterns, semantics, and more. LLMs leverage this understanding of language to perform specialized tasks.
- Machine learning (ML): A subfield of artificial intelligence that gives computers the ability to learn and complete tasks without being explicitly programmed, usually by identifying patterns and making predictions.
- Metadata: Data that provides information about other data. An example of metadata includes the resolution, size, and color saturation of images you take on your mobile phone.
- Multimodal model: A data model composed of multiple modalities (like text, audio, and video) that processes each modality separately. Multimodal models enable tools like AI-powered image generators.
- Natural language processing (NLP): The ability of a computer to understand written or spoken human language.
- Parsing: Identify the different elements that constitute a text, then assigning each element a grammatical and logical value.
- Personally Identifiable Information (PII): Information that permits the identity of the individual to whom the information refers to be reasonably inferred. PII includes things like names, addresses, telephone numbers, and more.
- Plugins: Software that extends the functionality of a large language model into other areas, like booking flights or performing mathematical calculations.
- Prompt: The instructions that set the task for an AI model.
- Prompt chaining: Using multiple prompts in a sequence to refine a request made to an AI model.
- Sentiment analysis: The process of analyzing text to identify the emotional tone (the “sentiment”) of the text.
- Training data: The dataset used in the early training of a machine learning model.