Skip to Main Content
Rider University Library

AI Glossary: Home

AI Glossary

  • AI (Artificial Intelligence): The ability of a digital machine to carry out tasks typically performed by intelligent beings, often involving computer vision, speech-to-text, and natural language processing. (Chiu et al.; US Dept of Ed)

  • AI Agents: Autonomous or semi-autonomous AI systems designed to operate within specific environments to accomplish goals, interacting dynamically and making decisions. (AI Index Report)

  • AI Education: The process of teaching students about AI, including its concepts, workings, ethical implications, and societal impact. Distinct from using AI in education. (AI Index Report; US Dept of Ed)

  • AI in Education: The usage of AI tools and systems to support and enhance the teaching and learning process. (AI Index Report; US Dept of Ed)

  • AI Literacy: The foundational understanding of AI—how it works, how to use it, and its ethical impacts—for non-expert users, enabling them to understand, use, monitor, and critically reflect on AI applications. (AI Literacy for Users; Chiu et al.; AI Index Report)

  • AI Model: Components of larger AI systems, consisting of computer code and numerical values ("weights") designed to accomplish specific tasks, like generating text or images. (US Copyright Office)

  • AI Tool Recognition: The ability to identify whether one is interacting with an AI tool or a human, which is becoming increasingly difficult due to AI's human-like capabilities. (AI Literacy for Users)

  • Algorithmic Literacy: Awareness and knowledge of algorithms, including trust and confidence in them, and understanding their potential for appreciation or avoidance. (Chiu et al.)

  • AlphaGeometry: A neuro-symbolic hybrid AI system developed by Google DeepMind that uses a language model trained on synthetic data to solve geometry problems at a high level. (AI Index Report)

  • Ambient AI Scribes: Technology that integrates Large Language Models (LLMs) to process physician-patient recordings, aiming to reduce the burden of clinical documentation. (AI Index Report)

  • ARC-AGI (Abstraction and Reasoning Corpus - Artificial General Intelligence): A challenging benchmark designed to test AI systems' general intelligence and ability to reason across broad domains. (AI Index Report)

  • Autonomy (Ethical): In the context of AI in education, refers to the capacity for self-governance and making informed decisions, which can be impacted by reliance on AI. (Strunk and Willis)

  • Benchmarks: Standardized tests or evaluation suites used to assess the capabilities and performance of AI systems across various tasks and domains. (AI Index Report)

  • Black Box Problem (AI): The phenomenon where the internal workings and decision-making processes of complex AI systems are difficult to interpret or explain, even by their developers. (US Copyright Office; US Dept of Ed)

  • C4 Common Crawl Dataset: A massive dataset of publicly available web data often used for training AI models, which has recently seen increased data use restrictions. (AI Index Report)

  • ChatGPT: A widely used generative AI chatbot, still the top-used AI tool in education. (Carnegie Learning)

  • Computational Resources ("Compute"): The computational power required to train and operate a machine learning model, directly influenced by model complexity and training data size. (AI Index Report)

  • Contamination (Benchmarks): Occurs when LLMs encounter test questions that were present in their training data, potentially inflating reported performance. (AI Index Report)

  • Critical Literacy (AI): The ability to think critically about AI models, their outputs, one's own usage, and the impact of AI tools, especially prominent due to AI's complexity and inscrutability. (AI Literacy for Users)

  • Data Augmentation: A technique that modifies real data (e.g., tilting or image mixing) to create new variations, expanding datasets while preserving essential characteristics. (AI Index Report)

  • Data Commons: Refers to the pool of publicly available web data used for training AI models. (AI Index Report)

  • Data Governance: Policies, procedures, and standards established by an organization to ensure the quality, security, and ethical use of data, particularly relevant for AI training data. (AI Index Report)

  • Data Privacy: An individual's right to confidentiality, anonymity, and protection of their personal data, including consent and information about data usage, crucial in AI systems. (US Dept of Ed; AI Index Report)

  • Deepfakes: Synthetic media, often videos or images, created using AI to manipulate or fabricate content, posing risks for misinformation and harassment. (AI Index Report)

  • Digital Equity: Ensures that all individuals and communities have the information technology capacity needed for full participation in society, democracy, and economy. (US Dept of Ed)

  • ELSI (Ethical, Legal, and Societal Implications): A broad term encompassing the ethical, legal, and societal issues arising from scientific and technological advancements, particularly relevant to AI. (AI Index Report Appendix)

  • Factuality: The degree to which AI-generated responses are accurate and truthful, an area of increasing concern and benchmark development. (AI Index Report)

  • Foundation Model: A type of AI model, typically large and trained on vast datasets, that can be adapted for a wide range of downstream tasks. (AI Index Report)

  • FURM (Fair, Useful, Reliable, Measurable) framework: A framework used by Stanford Health Care for evaluating and implementing AI tools in clinical settings. (AI Index Report)

  • Generative AI: AI systems capable of producing novel content, such as text, images, audio, or video, often from inputs like text prompts. (US Copyright Office; Strunk and Willis; AI Index Report)

  • Hallucinations (AI): Instances where AI models generate factually incorrect or nonsensical outputs, a significant challenge for reliability. (US Dept of Ed; AI Index Report)

  • HELM Safety (Holistic Evaluation of Language Models Safety): A benchmark for assessing the safety aspects of AI models. (AI Index Report)

  • Human in the Loop AI: A principle emphasizing that human judgment and control should remain central in systems where AI is integrated, especially in education. (US Dept of Ed)

  • Humanity’s Last Exam (HLE): A new, highly challenging benchmark comprising college-level, multimodal questions designed to test advanced AI capabilities and resist simple internet lookups. (AI Index Report)

  • Infectious Jailbreaks: A multi-agent vulnerability in multimodal large language model (MLLM) systems where compromising one agent can trigger a rapid, system-wide failure, spreading harmful behavior exponentially. (AI Index Report)

  • Inference Cost: The cost associated with querying or operating a trained AI model to generate outputs. (AI Index Report)

  • Inspectable AI: The characteristic of AI systems that allows educators to examine how AI models analyze student work and the rationale behind their recommendations. (US Dept of Ed)

  • Intelligent Tutoring Systems (ITS): AI-enabled learning systems designed to provide personalized, adaptive support and feedback to students as they learn, often in subjects like mathematics. (US Dept of Ed)

  • LLM (Large Language Model): A type of AI model trained on unimaginably vast quantities of text data to understand, generate, and process human language. (Strunk and Willis; New Yorker; AI Index Report)

  • Multimodal Models: AI models capable of processing and generating content in multiple formats, such as text, audio, images, and video. (AI Index Report)

  • Natural Language Processing (NLP): A field of AI that enables computers to understand, interpret, and generate human language. (Chiu et al.; AI Index Report)

  • Open-Weight Models: AI models where the underlying "weights" (parameters) are fully available to the public, allowing for independent modification and scrutiny. (AI Index Report)

  • Overridable AI: The ability for human educators to intervene and override decisions or recommendations made by an AI system when they disagree with its logic. (US Dept of Ed)

  • Prompt: A common type of input, often in text form, that communicates the desired features of an output to an AI system. (US Copyright Office)

  • Prompt Engineering: The practice of crafting prompts that are optimized to elicit a desired result from a generative AI system. (US Copyright Office)

  • Retrieval-Augmented Generation (RAG): An AI approach that integrates LLMs with retrieval mechanisms to enhance response generation by first retrieving relevant information from external sources. (AI Index Report)

  • Robots.txt: A file that specifies rules for web crawlers, increasingly used by websites to restrict data scraping for AI training, impacting the "data commons." (AI Index Report)

  • Self-Reflective Mindsets (AI): A component of AI competency that involves students consistently evaluating their AI knowledge to stay up-to-date, linked to confidence and lifelong learning. (Chiu et al.)

  • Synthetic Data: Data generated by AI models themselves, explored as a solution to potential data shortages for training AI, though with concerns about quality and fidelity. (AI Index Report)

  • Transparency (AI): Encompasses open sharing of AI development choices (data sources, algorithmic decisions) and how systems are deployed and monitored, often including explainability. (AI Index Report)

  • Turing Test: A test proposed by Alan Turing in 1950 to evaluate a machine's ability to exhibit human-like intelligence, now considered largely surpassed by advanced LLMs. (AI Index Report)

  • Zuckerberg Parenthesis: A metaphorical period of history inaugurated by Facebook, where social media ruled, preceding the significant threat of AI to the conversational internet. (New Yorker)