Princeton IT Specialist Flashcards
ChatGPT
Artificial Intelligence (AI)
The simulation of human intelligence by machines. AI allows computers to perform tasks that normally require human thinking, like understanding language or recognizing images.
AI can help us translate languages instantly.
Using AI to analyze and identify communication trends to optimize the timing and delivery of messages; Applying AI tools like ChatGPT or Copilot to draft or enhance communications for different internal channels; Leveraging AI to automate routine tasks, such as curating content for SharePoint or MS Teams channels.
Machine Learning (ML)
A type of AI where computers learn from data to improve their performance over time. The more data the computer gets, the better it becomes at tasks.
Machine learning helps language models understand grammar rules.
Using ML algorithms to measure the effectiveness of different communication channels and suggest improvements; Employing ML tools to categorize and prioritize new information coming from Enterprise BI&T; Analyzing feedback and refining communication strategies based on ML-driven insights.
Language Model
A program that can understand and generate human language. It can read text, answer questions, or even hold conversations.
A language model can help you practice writing sentences in English.
Utilizing language models (e.g., ChatGPT) to generate draft content for newsletters, Town Hall presentations, or internal training materials; Implementing language models for automated responses or assistance within MS Teams channels; Enhancing the quality and consistency of internal communications by proofreading content using language models.
Training Data
-The information that a language model learns from. It usually includes a lot of text from books, websites, and other written sources.
-The training data helps the language model understand how people use words.”
-Collecting and organizing past communications as training data to help refine automated content generation or analysis tools.
-Using internal documentation and stakeholder feedback as training data to customize AI tools for specific communication needs.
-Continuously updating training data to ensure the relevance of information shared across channels like SharePoint or Viva Engage.
Natural Language Processing (NLP)
-A field of AI focused on helping computers understand, interpret, and generate human language.
-“NLP is used in voice assistants like Siri and Alexa to understand what we say.”
-Using NLP tools to analyze large amounts of text data (emails, internal reports) to extract key insights for the communications strategy.
-Applying NLP to identify the most frequently discussed topics within the organization and tailor communications accordingly.
-Utilizing NLP to help automate content tagging and categorization in SharePoint.
Prompt
Definition: The text or question given to a language model to generate a response.
Example sentence: “You can use a prompt like ‘Tell me a story about a dragon’ to see how the language model creates a story.”
Crafting prompts for AI tools like ChatGPT to generate draft content for newsletters, spotlight series, or technical training materials.
Creating prompts to assist in content generation for presentations, ensuring clarity and alignment with BI&T’s goals.
Using structured prompts to guide AI in extracting relevant information from long documents.
Algorithm
Definition: A set of instructions that a computer follows to solve a problem or complete a task.
Example sentence: “The algorithm in a language model helps it decide what word should come next in a sentence.”
Employing algorithms to recommend the best time for sending important announcements to maximize engagement.
Using algorithms to automate the sorting of information collected from different business areas, ensuring that the most relevant content is highlighted.
Developing algorithms to prioritize urgent communications for immediate dissemination.
Context
Definition: The background information or surrounding words that help a language model understand what a sentence means.
Example sentence: “In the sentence ‘He saw the bank,’ context helps the model know if ‘bank’ means a riverbank or a financial institution.”
Providing sufficient context for communications by tailoring messages to different audiences within the company.
Using the context of past projects or accomplishments to frame updates and highlights shared with stakeholders.
Ensuring that AI-generated content is relevant by including specific context when prompting tools for content creation.
Response
Definition: The text that the language model generates after being given a prompt.
Example sentence: “If you ask the language model ‘What is the capital of France?’, the response would be ‘Paris.’”
Preparing responses for anticipated questions during presentations or internal communications.
Using AI-generated responses to handle routine inquiries on MS Teams or SharePoint forums.
Refining responses based on feedback gathered from various channels to improve future communications.
Bias
Definition: When a language model gives answers that are influenced by the data it was trained on, sometimes leading to unfair or one-sided results.
Example sentence: “If a language model was trained only on stories from one country, it might have a bias toward that country’s culture.”
Being mindful of bias in AI-generated content to ensure inclusivity in internal communications.
Reviewing content for any biases before amplifying messages to the organization.
Training staff on how to recognize and avoid bias in content creation, especially when using AI tools.
Fine-Tuning
Definition: The process of training a language model further on a specific type of data to make it better at a particular task.
Example sentence: “Fine-tuning a language model with medical texts can help it answer health-related questions.”
Fine-tuning AI tools to better understand the specific language and tone preferred by the organization.
Continuously updating content strategies based on the effectiveness of past communications.
Fine-tuning language models to capture the unique voice of the BI&T department.
Token
Definition: A piece of text, such as a word or part of a word, that a language model uses to understand language.
Example sentence: “The sentence ‘I love dogs’ has three tokens: ‘I,’ ‘love,’ and ‘dogs.’”
Understanding how tokens are used in AI tools to improve prompt creation for content generation.
Managing communication costs when using AI models by keeping prompts concise.
Using token-based analysis to measure content length for different communication channels.
Parameter
Definition: The internal settings in a language model that help it generate responses. More parameters usually mean the model can understand language better.
Example sentence: “A language model with millions of parameters can create very realistic conversations.”
Adjusting AI tool parameters to prioritize clarity and consistency in automated content generation.
Setting parameters for AI tools to focus on specific keywords or topics within communications.
Using parameter adjustments to control the tone and style of AI-generated content for different channels.
Inference
Definition: The process of the language model using what it has learned to generate a response to a new prompt.
Example sentence: “During inference, the language model figures out what answer to give based on the prompt.”
Using inference tools to predict the most relevant information to share with stakeholders.
Applying inference techniques to identify potential areas of interest based on past engagement patterns.
Automating content suggestions for SharePoint or Teams based on inferred relevance.
Chatbot
Definition: A computer program that uses a language model to have conversations with people.
Example sentence: “A chatbot can help answer your questions when you visit a website.”
Implementing a chatbot within MS Teams to assist staff in finding relevant resources or answering common questions.
Using chatbots to guide employees through using internal tools like Viva Engage or SharePoint.
Leveraging chatbots for quick dissemination of updates, announcements, or reminders.
Pre-training
Definition: The initial phase where a language model learns from a large amount of text data before being used for specific tasks.
Example sentence: “Pre-training allows the language model to understand many words and phrases before it starts answering questions.”
Ensuring that any AI tools used for content creation are pre-trained on internal communication styles.
Customizing AI tools with pre-training to align with Research BI&T’s tone and terminology.
Using pre-training to improve the quality of content created for Town Hall presentations or newsletters.
Human-in-the-Loop
Definition: A method where humans help improve the language model’s answers by providing feedback or corrections.
