Glossary of ChatGPT, Prompt Engineering, and Artificial Intelligence

The simulation of human intelligence processes by machines, especially computer systems, including learning, reasoning, and self-correction.

Example: Narrow AI applications like facial recognition software and virtual assistants such as Siri or Alexa.

A type of artificial intelligence that creates images from textual descriptions, leveraging deep learning techniques to interpret and visualize concepts.

Example: AIGC tools like DALL-E, Midjourney, and Leonardo AI

A mathematical framework or algorithm that is trained on data to perform specific tasks, such as classification, regression, or generation. AI models utilize various techniques, including machine learning and deep learning, to learn patterns from data and make predictions or generate outputs.

Example: GPT-4, BERT, DALL-E

A type of artificial intelligence that produces written content based on prompts or instructions, utilizing language models to generate coherent and contextually relevant text. It create articles, stories, or answers to questions based on user input.

Example: ChatGPT functions as an AI text generator (although it is also multi-modal, capable of generating images and voice responses).

An AI Video Generator is a tool that creates videos from text prompts or templates using AI technology. In genreal, these tools can generate animations, edit videos, or even create lifelike avatars that speak scripted content.

Example: Runway, InVideo, Pika Labs

Content created by artificial intelligence models, including text, images, audio, and video, based on input prompts or instructions.

Example: Articles, images, and even videos created by ChatGPT based on user-provided prompts.

Reference: What is AIGC? Artificial Intelligence Generated Content Explained Using ChatGPT

A sequence of instructions or rules that a system or a program (e.g. an AI model) follows to process data and make decisions. Algorithms are used in many fields to process data, automate calculations, or guide decision-making processes.

Example: ChatGPT uses a next-word prediction algorithm, which analyzes the context of a sentence and predicts the most likely next word based on patterns it learned from training data.

A reinforcement learning AI developed by DeepMind that plays the board game Go, known for its ability to defeat top human players through advanced strategies and self-play training.

Link: https://deepmind.google/research/breakthroughs/alphago/

A specific keyword or phrase included in a prompt to guide an AI’s focus and help shape its response around desired themes or tones. Anchor words give the AI subtle cues to stay on topic or emphasize particular ideas.

Example: In a prompt asking for business ideas, anchor words like “creativity” and “future” lead the AI to generate forward-thinking concepts.

A set of rules that allows different software applications to communicate and work together. In AI, APIs enable models like ChatGPT to be integrated into other platforms or tools.

Example: Content platforms can use ChatGPT’s API to help users draft articles and social media posts directly within their own applications.

A prompt technique where the AI model is encouraged to seek clarification from users before providing an answer. This approach enhances response accuracy and relevance by enabling a back-and-forth dialogue, allowing users to specify their needs.

Example Prompt: “Here are the requirements for this article: [Requirement A], [Requirement B], and [Requirement C]. Please generate the article based on these. Feel free to ask any questions if you need clarification”

Reference: What is ‘Ask Before Answer’? Prompt Technique to Ensure Accurate and Specific Responses from ChatGPT

An approach that combines human intelligence with AI to enhance, rather than replace, human decision-making and problem-solving. Augmented intelligence supports users by providing insights and recommendations while leaving final decisions to people.

Example: In healthcare, an AI system might analyze patient data and suggest potential diagnoses, but doctors make the final call.

A technique where the AI generates follow-up prompts automatically based on the user’s initial input, guiding the conversation or task without requiring manual prompts for each step.

Example: If a user asks about planning a trip, auto-prompting might follow up with, “Would you like suggestions on accommodations or activities?”

The use of technology, such as artificial intelligence, to perform tasks without human intervention. This is especially helpful for repetitive or rule-based processes, as it saves time and reduces errors.

Example: An AI-powered customer service bot automatically answers common inquiries without human intervention.

(Refer to Google’s Bard)

A tendency in AI models to produce responses that reflect certain preferences, stereotypes, or inaccuracies based on patterns in training data. Bias can affect the fairness and objectivity of an AI’s outputs.

Example: AI tool intended to streamline the hiring process might be biased against applicants of a certain age group.

Extremely large and complex datasets. Big data is used to train AI models to enable them to identify patterns, make predictions, and improve performance over time.

Example: ChatGPT is trained on big data that includes a diverse range of internet text, books, articles, and other resources.

A prompting technique where the user guides the AI to break down its reasoning step-by-step to solve complex tasks or answer detailed questions accurately. Magic phrases like “let’s go step-by-step” or “let’s do it step-by-step” are often employed to achieve this.

Example Prompt:  “How does photosynthesis work? Explain it step-by-step.”

Reference: What Is ‘Chain of Thought’ Prompting? Step-by-Step Reasoning to Improve ChatGPT’s Outputs

A record of previous interactions within a conversation. This record allows both the system and the user to refer back to earlier messages.

Example: AI models may record chat history to maintain coherence in subsequent conversations, to provide more personalized experience.

A software application that simulates conversation with users, often through text or voice. Chatbots can be rule-based (following predefined scripts) or AI-powered. They handle tasks like answering questions, providing information, and assisting with customer service.

Example: Many websites use chatbots to answer commonly asked questions.

An AI language model developed by OpenAI and a popular AIGC (AI-Generated Content) tool that generates human-like text responses based on user prompts. ChatGPT assists with various tasks, such as answering questions, drafting text, and engaging in conversational dialogue.

Link: https://chatgpt.com/

Reference: What Can ChatGPT Do For Me?

A subscription plan offered by OpenAI that provides paid users with enhanced access to ChatGPT, including faster response times and priority access to new features and improvements.

Link: https://openai.com/chatgpt/pricing/

A feature that enables users to interact with ChatGPT using spoken language, allowing for more natural and engaging conversations. This capability uses advanced speech recognition and synthesis technology to convert spoken input into text and generate spoken responses.

Link: https://openai.com/index/chatgpt-can-now-see-hear-and-speak/

An AI language model developed by Anthropic, designed to assist with a variety of text-based tasks, similar to ChatGPT. Claude AI focuses on safe, reliable, and interpretable responses, and is used for tasks such as answering questions, drafting content, and engaging in conversation.

Link: https://claude.ai/

[For AI in general] Context is the information surrounding an interaction that provides meaning and relevance to the interaction, helping the system or application (e.g. ChatGPT) understand and respond appropriately based on previous inputs and user intent.

Example: In ChatGPT, context includes the history of messages exchanged during a session.

[For prompt engineering] Context refers to the surrounding information or details provided in a prompt that help guide the AI model’s understanding and response. This includes any relevant background information, specific instructions, or examples.

Example: In the following prompt, the first sentence provides context: “I am a cat owner of five furry babies, living in a small apartment with limited space. I am about to bring in a dog to join the family…”

A prompt engineering technique where background information, context, or other additional details are embedded into a prompt to guide the AI’s response. This helps the AI understand the user’s intent more clearly, and will lead to more relevant and accurate outputs.

Example: The following prompt provides context before the main task: “My daughter is 10 years old and has autism. Please suggest activities she might enjoy.”

The maximum amount of text (measured in tokens) that an AI model can process in a single interaction. The context window includes both the user’s input and the AI’s response, and once it’s full, older information may be “forgotten” as new text is added.

Example: ChatGPT-4 has a context window of up to 8,000 or 32,000 tokens, depending on the model version.

A prompting technique where relevant context or background information is included in the prompt to help the AI understand specific situations or complex requests, so that it can generate responses that are more accurate and aligned with the user’s needs.

Example: Instead of asking, “What are some good travel tips?” contextual information can be added, like: “I’m traveling to Japan in winter with two kids—what are some good travel tips for a family?

A type of artificial intelligence designed to simulate human-like conversations through text or voice. Conversational AI uses natural language processing (NLP) to understand user inputs and respond contextually, making it useful for applications like customer support, virtual assistants, and chatbots.

Example: ChatGPT and virtual assistants like Siri or Alexa

(Refer to Microsoft’s Copilot)

A technique that encourages the AI to generate imaginative, unique, or unconventional responses by framing prompts in an open-ended or exploratory way. This approach is often used for storytelling, brainstorming, and artistic tasks where originality is valued.

Example: A creative prompt might be, “Imagine a futuristic city where people communicate only through colors. Describe a day in this city.”

A tailored version of the GPT model that can be fine-tuned or configured to meet specific user requirements or applications. Custom GPTs allow the users to create customized AI experiences that align with their unique needs and contexts, including adjustments in personality, tone, and domain-specific knowledge.

Example: Guru GPT that allows ChatGPT to search across the user’s apps, docs, and chats.

Reference: 
What Is A Custom GPT for ChatGPT? — What You Need to Know About Custom GPTs (Part 1)
How To Create A Custom GPT for ChatGPT? — What You Need to Know About Custom GPTs (Part 2)

DALL-E is an AI model developed by OpenAI that generates images from text descriptions, allowing users to create visuals based on specific prompts. ChatGPT’s image generation capability is powered by DALL-E. It enables users to request images directly within the ChatGPT interface.

Link: https://openai.com/index/dall-e/

The process of analyzing large datasets to discover patterns, trends, and useful information. Data mining is widely used in AI and business to gain insights, make predictions, and support decision-making.

Example: E-commerce companies use data mining to analyze customer purchasing patterns, helping them recommend products and personalize marketing efforts.

The practice of safeguarding personal or sensitive information to ensure it is collected, processed, and stored securely and used responsibly. Data privacy measures help protect individuals’ information from unauthorized access or misuse.

Example: ChatGPT avoids retaining personal information from conversations to align with data privacy best practices.

A collection of structured information used to train, validate, or test AI models. Datasets are essential for helping AI learn patterns, make predictions, and generate responses based on examples.

Example: ChatGPT is trained on vast text datasets containing books, articles, and web content.

A subset of machine learning that uses layered neural networks to analyze complex data and make predictions. Deep learning enables AI models to recognize patterns in large datasets. It powers applications like image recognition, language processing, and natural language understanding.

Example: ChatGPT uses deep learning techniques to learn patterns in text data. This enables it to generate human-like responses during interactions.

A technology that uses AI to create synthetic media, such as images, videos, or audio. It is capable of convincingly mimicking faces, voices, and movements with high realism.

Example: Deepfake videos are created by superimposing an actor’s face onto another person’s body, making it appear as if the actor is performing actions they didn’t actually perform.

A technology that uses AI to identify and interpret human emotions from data sources such as facial expressions, voice tone, or text.

Example: Emotion recognition commonly applied in customer service field to understand and respond to users’ emotional states.

A prompting technique where specific examples are included in the prompt to guide the AI’s response style or format. This approach helps the AI understand the desired output by providing clear reference points.

Example: When asking ChatGPT to imitate a writer’s style, the user can provide sample writings from that writer as examples.

Reference: What Are Zero-Shot, One-Shot, and Multiple-Shot Prompting, And Why Is It A Prompt Engineering Technique You Must Know

A technology that uses AI to identify or verify individuals based on their facial features.

Example: Face recognition is used by smartphones to unlock the devices.

A prompting technique where the user provides a few examples within the prompt to guide the AI’s response, to help the AI understand the desired output and its format, tone, or style.

Example: To generate a set of exam questions, a user might provide past-year exam questions for the AI’s reference.

Reference: What Are Zero-Shot, One-Shot, and Multiple-Shot Prompting, And Why Is It A Prompt Engineering Technique You Must Know

A prompting technique where the user provides a sentence or phrase with missing parts for the AI to complete. This approach is useful when the user needs help finding suitable words, information, or ideas.

Example Prompt: “Complete this sentence: The benefits of exercise include: 1. Improved mood, 2. Increased energy, 3. ___,”

Reference: What is Fill-in-the-Blank Prompting? A Brainstorming Technique to Complete Your Incomplete Thoughts

A prompting technique that involves using additional prompts to refine or expand upon the AI’s initial responses. This helps guide the AI more effectively in multi-step tasks or complex questions.

Example: After the AI suggests the general benefits of exercise, a follow-up prompt that narrow down the focus might be, “Elaborate on how exercise boosts mental well-being.”

(Refer to Google’s Gemini)

A prompt engineering technique where an AI model is requested to first generate relevant knowledge based on a task or question before using that information to formulate a response. This two-step process enhances the model’s ability to provide accurate and contextually relevant answers by leveraging its generated knowledge.

Example:Apply this technique by issuing two successive prompts: Prompt #1 requests ChatGPT to explain the rules of Facebook advertising; Prompt #2 instructs ChatGPT to write a Facebook ad that adheres to the knowledge generated in Prompt #1.

Reference: Advanced Prompt Engineering Technique 2: Generated Knowledge Prompting

A type of artificial intelligence that creates new content—such as text, images, audio, or video—based on patterns learned from existing data.

Example: AI tools like ChatGPT, DALL-E, Midjourney, Suno

Reference: What is AIGC? Artificial Intelligence Generated Content Explained Using ChatGPT

Bard was Google’s conversational AI tool, now succeeded by the Gemini model suite, designed for more accurate, context-aware responses. It assists users with questions, tasks, and information retrieval.

Link: https://bard.google.com (will be redirected to Gemini)

Gemini is Google’s suite of advanced AI models, designed to power conversational tools and enhance various applications with improved language understanding and multimodal capabilities. It is the successor to Google’s Bard.

Link: https://gemini.google.com/app

A type of AI model developed by OpenAI that generates human-like text based on user input. GPT models are trained on large datasets and use transformer architecture to understand and generate language.

Example: ChatGPT, based on GPT technology, can answer questions, assist with writing, and engage in conversation by generating responses from learned language patterns.

In AI, hallucination refers to a situation where a model generates information or details that are not based on real data, facts, or context. These responses may seem plausible but are actually fabricated by the model.

Example: ChatGPT might hallucinate by providing a fictional statistic or misattributing a quote.

A type of artificial intelligence that combines different AI techniques, such as rule-based systems, machine learning, and deep learning, to achieve more flexible and effective problem-solving.

Example: A customer service system using hybrid AI, for example rule-based logic for simple queries, and machine learning to handle open-ended questions.

A technology that enables AI to identify and classify objects, people, or other elements within images.

Example: Image recognition is widely used in applications such as facial recognition, autonomous driving, and medical imaging.

Data or information provided to a system or application to process and generate a response or output. In the context of AI interactions, input usually refers to user prompts or commands that guide the model’s response.

Example: Prompts for ChatGPT

A technique where prompts are refined and adjusted in multiple rounds to gradually improve the AI’s response. Iterative prompting allows users to clarify instructions or ask follow-up questions, and this will help enhancing the accuracy and relevance of the output.

Example: If ChatGPT’s initial answer is too general, a user might refine their prompt by adding specifics, like “Please focus on examples for cat owners.”

A collection of structured information or content that an AI system can reference, to provide accurate answers and support decision-making.

Example: An AI tutor uses a knowledge base to answer questions on specific academic topics.

A specific point in time up to which an AI model, particularly a large language model, has been trained on data. This date represents the final point of data available to the model, restricting its knowledge of events or updates beyond that time.

Example: The knowledge cutoff date of ChatGPT-4o is October 2023.

Reference: What Are ChatGPT’s Knowledge Cutoff Dates and Why They Matter

An AI system trained to understand, generate, and analyze human language by learning patterns from large amounts of text data.

Example: ChatGPT is a language model designed to generate human-like responses.

An advanced type of AI model trained on vast amounts of text data has a large number of parameters (often billions), allowing it to understand and generate human-like language. LLMs can perform a wide range of language tasks, including answering questions, summarizing information, and engaging in conversation.

Example: ChatGPT is powered by an GPT-4, Google’s Gemini is powered by Gemini 1.5, Claude AI is powered by Claude 2—all of which are LLMs.

Reference: What is a “Large Language Model”? LLM Explained Using ChatGPT

The delay between a user’s input and a system or application’s response, which can affect their responsiveness. Latency is often measured in milliseconds.

Example: The average latency of ChatGPT for most text-based queries is around 1.5 seconds per response.

A prompting technique that starts with simple, low-effort questions or tasks and gradually progresses to more complex ones. The aim is to facilitate understanding and engagement.

Example: A teacher might first ask ChatGPT to name animals before moving on to more challenging questions about animal habitats.

A generative AI platform designed for creating and manipulating high-quality visual content, including images, animations, and 3D textures. By leveraging advanced algorithms and machine learning techniques, Leonardo AI offers a user-friendly interface with a wide array of features, including image generation, real-time canvas, 3D texture generation, motion, and custom model training.

Link: https://leonardo.ai/

A subset of artificial intelligence that focuses on the development of algorithms and statistical models. Machine Learning enables computers to learn from and make predictions based on data without being explicitly programmed.

Example: Fraud detection, spam filtering, and image recognition

A component of artificial intelligence systems that allows them to retain information from previous interactions. Memories enable more personalized and contextually relevant responses in future conversations.

Example: ChatGPT uses memory to remember certain user preferences and conversation history.

An advanced AI-driven assistant developed by Meta (formerly Facebook) that utilizes the LLaMA (Large Language Model Meta AI) technology. It is designed to enhance user interactions across Meta’s platforms, including Facebook, Instagram, WhatsApp, and Messenger, by providing personalized assistance, answering queries, and facilitating content creation.

Link: https://www.meta.ai/

A subfield of machine learning that focuses on designing algorithms that can learn how to learn, to improve their learning efficiency and performance on new tasks by leveraging previous experiences.

Example: Few-Shot Learning – AI models learn to recognize new objects from just a few examples by using knowledge gained from previously learned tasks.

An AI-powered assistant integrated into Microsoft 365 applications like Word, Excel, and Outlook to enhance productivity. It assists users by generating text, analyzing data, creating presentations, and automating repetitive tasks based on prompts within these tools.

Link: https://copilot.microsoft.com/

An AI-powered image generation tool that creates high-quality artwork and visuals from textual prompts, allowing users to explore creative possibilities in various artistic styles.

Link: https://www.midjourney.com/

A term that refers to the capability of AI systems to process and integrate various types of input data, including text, images, audio, and video, for richer interactions and enables AI to perform tasks that require understanding across different modalities.

Example: ChatGPT’s multi-modal capabilities enable it to process text, audio, and image inputs.

Reference: What is Multimodal AI? Understanding ‘Multimodality’ with ChatGPT

(Refer to Few-Shot Prompting)

A field of artificial intelligence that focuses on the interaction between computers and humans through natural language, training the machines to understand, interpret, and generate human language in a meaningful way.

Example: NLP is used in applications like chatbots, sentiment analysis, and language translation services

A prompting technique that involves specifying what should not be included in the output, helping to guide the AI model away from undesired responses or content.

Example Prompt: “Do not include any violence or dark themes.”

A computational model inspired by the human brain, consisting of interconnected layers of nodes (neurons) that process data and learn from it, allowing the model to recognize patterns and make decisions.

Example: Neural networks are widely used in image and speech recognition tasks

An artificial intelligence research organization focused on developing and promoting friendly AI for the benefit of humanity, known for creating advanced AI models and technologies. OpenAI is the developer of ChatGPT, DALL-E, and several other innovative projects.

Link: https://openai.com/

The information or content generated by a system or application in response to a given input. In AI, outputs can take various forms such as text, images, or audio.

Example: In a conversation with ChatGPT, the output consists of the generated text responses based on the user’s queries.

The process of adjusting and fine-tuning the results generated by a system or an application, to ensure that the output meets desired quality standards and aligns with user expectations.

Example: In an AI text generation model, output calibration might involve tweaking the response to make it more coherent or relevant to the user’s query.

A variable or setting that influences the behavior and performance of a system or an application. Parameters are often used to define the architecture of the system or application, or to control learning processes during training.

Example: ChatGPT 4 is purportedly trained on over a trillion parameters

The ability of a system to identify and categorize patterns within data, allowing it to make predictions or decisions based on those patterns.

Example: In image processing, pattern recognition enables systems to detect objects, faces, or other features in photos and videos.

An AI-powered research and conversational search engine developed by Perplexity, Inc., designed to answer queries using natural language processing. The platform supports multiple large language models, including GPT-4 Turbo, Claude 3, Sonar Models, and Mistral Large. It synthesizes information from various sources and provides responses with inline citations, making it a versatile tool for information discovery.

Link: https://www.perplexity.ai/

Reference: Perplexity as a Key Competitor of ChatGPT: Dealing with AI Models’ Knowledge Cutoff Dates

[For AI] A defined character or identity that an AI model adopts in interactions with users, shaping its tone, style, and manner of communication to suit specific contexts or user preferences. Assigning persona to AI models is an important prompt engineering technique.

Example: A virtual assistant designed for elderly users may adopt a warm and patient persona.

A prompting technique that tailors the AI’s responses based on a defined persona or role, guiding the model to adopt specific characteristics, knowledge, or style in its outputs.

Example Prompt: “Imagine you are a manager at a multinational company with 10 engineers reporting to you…”

The process of tailoring content, responses, or experiences to individual user preferences, behaviors, or characteristics to enhance engagement and relevance.

Example: In ChatGPT, personalization allows the model to remember a user’s preferences and topics of interest across sessions.

The initial phase of training a machine learning model, particularly in natural language processing (NLP). During this phase, the model learns from a large dataset to grasp language patterns, structures, and general knowledge.

Example: In the context of ChatGPT, pre-training involves exposing the model to diverse text data to develop a foundational understanding of human language

A statistical technique used in machine learning to create models that predict future outcomes based on historical data, identifying patterns and relationships within the data.

Example: In marketing, predictive modeling can be used to forecast customer behavior.

A piece of text or instruction provided to an AI model to elicit a specific response or output, guiding the model’s generation process.

Example: In ChatGPT, a user might enter the prompt “Explain the benefits of meditation,” to instruct it to generate a detailed response on the topic.

The practice, or a set of techniques, of designing and refining prompts to improve the quality and relevance of the responses generated by AI models.

Example: Chain-of-thought prompting, ask-before-answer prompting, being clear and specific

Reference: What is Prompt Engineering and How to Improve Your ChatGPT Prompts

A structured approach to creating prompts that guide the interaction with an AI model to ensure consistency and effectiveness in generating desired outputs.

Example: RICE framework, CORE framework, IDEA framework

Reference: What Is a Prompt Framework? Top 7 Prompt Frameworks That Will Help You Heighten the Accuracy of ChatGPT’s Answers

A type of prompt designed to elicit specific information or responses from an AI model by posing direct questions. These prompts guide the AI by clearly stating what information is being sought, allowing for more precise and relevant outputs.

Example Prompt: “What are the key benefits of renewable energy?”

The capability of a system to process data and produce outputs instantly or within a very short time frame, for immediate responses to user inputs or external events.

Example: In conversational AI like ChatGPT, real-time processing allows the model to generate responses to user queries within seconds.

A type of machine learning where an agent (e.g., AlphaGo) learns to make decisions by taking actions in an environment to maximize cumulative rewards, based on feedback received from its actions.

Example: In a gaming context, a reinforcement learning model might learn to play chess by receiving rewards for winning games and penalties for losing.

The output generated by a system or an application (e.g. an AI model) in reaction to a user’s input or query. Responses can vary in form, including text, images, or audio.

Example: In ChatGPT, a user might ask, “What are the benefits of exercise?” and the response would be a detailed explanation of the health advantages.

A technique in prompt engineering where users provide the desired output or result they want the AI model to achieve, and then ask the model to generate the corresponding input or context that would lead to that output. This effectively acts as a form of reverse engineering the reference input.

Example: A programmer uploads a series of programming code to ChatGPT and then requests it to suggest a prompt that he can use to generate similar code. His prompt could be: “Based on the uploaded code snippets, please suggest a prompt that I can use to generate similar programming code.”

Reference: What Is Reverse Prompting? How to Reverse Engineer Prompts with ChatGPT

RICE prompt framework stands for Role, Instructions, Context/Constraints, and Examples. It is designed to include key prompt elements to help users create effective prompts for AI models to achieve the desired output.

Example Prompt: [Role] Imagine you are a veterinarian. [Instructions] Create a care guide for first-time cat owners. [Context/Constraints] This guide should be suitable for cat owners living in cities with limited living spaces. [Examples] For instance, include tips like: “Place the litter box in a quiet, accessible location.”

Reference: Want the Best Prompt Framework? We Recommend The RICE Prompt Framework

A training approach that combines reinforcement learning and human feedback to improve AI models, enabling them to learn more effectively by incorporating human evaluations into their learning process.

Example: RLHF is used to refine responses of AI models by training them based on feedback from users and evaluators.

A versatile platform that utilizes artificial intelligence to create high-quality videos, images, and audio. Runway offers features such as text-to-video generation, allowing users to produce videos by simply describing their desired content through text prompts.

Link: https://runwayml.com/

A technique where an AI model is encouraged to generate its own questions based on a given context or topic, to guide its responses and enhance the relevance of the generated content.

Example Prompt: “Discuss how to introduce a kitten into a new home. Generate relevant sub-questions that explore different aspects of this topic. Answer each of them in detail.”

Reference: Advanced Prompt Engineering Technique 1: Self-Ask Prompting

A temporary period during which a user engages with an application or service to allow for the exchange of information. The context of user engagement is maintained throughout the interactions during this period and is typically removed at the end of the session for privacy and security reasons.

Example: In ChatGPT, a session refers to the ongoing conversation between the user and the model. During such a period, ChatGPT retains context and remembers previous inputs until the session ends.

A unique identifier generated for a specific user session in an application, used to maintain state and track user interactions over time to ensure continuity during a conversation.

Example: ChatGPT uses session tokens to track the context of the ongoing conversation. This enables it to provide coherent and relevant responses based on the ongoing interactions.

Sora is a text-to-video AI model developed by OpenAI that generates videos from textual prompts, transforming written content into engaging visual stories. It offers advanced features such as the ability to create complex scenes with multiple characters, real-time collaboration capability, and an understanding of physical world dynamics. Currently, access to Sora is limited to select testers and creative professionals as OpenAI evaluates its functionality and addresses safety concerns.

Link: https://openai.com/index/sora/

A technology that translates the conversion of spoken language into text by analyzing audio input and identifying words and phrases, supporting interaction between humans and machines through voice commands.

Examples: Speech recognition is implemented in ChatGPT’s Voice feature, IBM Watson Speech to Text, Interactive Voice Response (IVR)

An approach in machine learning and artificial intelligence that involves training algorithms using labeled datasets to classify data or predict outcomes accurately. The system learns from examples that have correct answers, uses these examples to recognize patterns and make predictions about new information.

Example: To train an AI model to differentiate between cats and dogs, the developers show it many pictures of cats and dogs, telling it which is which. The system will learn to identify new pictures as either cats or dogs based on what it has learned.

[In AI responses] A parameter that controls the randomness of an AI model’s responses (i.e., how random the outputs are). Lower values typically produce more deterministic and focused responses, while higher values generate more diverse and creative outputs.

Example: In ChatGPT, setting the temperature to a low value (e.g., 0.2) will yield responses that are more predictable and conservative, whereas a high temperature (e.g., 0.8) will result in more varied and imaginative replies.

A conversation session between a user and an AI model that lasts for a limited duration or until the session is closed, where the context and memory of the interaction are not retained after the session ends.

Example: Temporary chat offered by ChatGPT that does not remember any details from this conversation once it’s over, to ensure user privacy and data security

Reference: What Is ChatGPT’s Temporary Chat? A Guide to Enhancing Your Privacy

A unit of text that the AI model processes, which can be as short as one character or as long as one word, depending on the language and context. Tokens are the building blocks of the input and output for AI models. They influence how they understand and generate language.

Example: In the context of ChatGPT, a sentence like “I love cats” might be broken down into tokens like “I,” “love,” and “cats,” each representing a distinct unit of meaning.

The maximum number of tokens that an AI model can process in a single request, including both input and output tokens. This limit affects the length of the text that can be sent to the model and the response it generates.

Example: In ChatGPT, if the token limit is set to 4096 tokens, the combined length of the user’s input and the model’s response cannot exceed this limit.

The process of breaking down text into smaller units called tokens, which can be words, phrases, or characters, so that a system can analyze and understand the structure and meaning of the text.

Example: In the sentence “The cat sits,” tokenization would result in three tokens: “The,” “cat,” and “sits.”

The collection of examples used to train a machine learning model, consisting of input-output pairs that the model learns from to make predictions or generate responses.

Example: A wide range of text data, including books, articles, and websites, is used as training data to train ChatGPT to understand language patterns and context.

A type of neural network architecture that revolutionized natural language processing by using self-attention mechanisms to process data in parallel, improving efficiency and performance in tasks like translation and text generation.

Example: The Transformer architecture is the foundation for AI models like BERT and GPT.

A type of machine learning where an algorithm is trained on data without labeled responses, This enables the model to independently discover patterns, groupings, or structures within the data. Such an approach is beneficial for exploring datasets and uncovering hidden relationships.

Example: In anomaly detection work, unsupervised learning is utilized to identify unusual patterns or outliers in datasets.

An AI-powered system designed to interact with users through natural language, providing assistance and support in various tasks. Virtual agents can operate via text or voice and are often integrated into websites, messaging apps, and customer support systems.

Example: Chatbots integrated into websites to answer frequently asked questions from customers and provide 24/7 instant support.

An AI-powered system that recognizes and processes voice commands, enabling users to perform tasks, access information, and control devices through spoken language.

Example: Amazon Alexa, Google Assistant, Apple Siri

The strategy of formulating prompts without providing any prior examples for the AI model’s reference. This approach relies on the model’s inherent understanding and general knowledge to generate an appropriate response.

Example Prompt: “Tell me about Newton’s First Law.” (no example is given to guide the AI)

Reference: What Are Zero-Shot, One-Shot, and Multiple-Shot Prompting, And Why Is It A Prompt Engineering Technique You Must Know