Definitions of Terms

Below are key terms and concepts in generative AI that are relevant to health professions education:

• AI literacy: The knowledge and skills required to effectively, responsibly, and critically evaluate and use AI tools, a core competency for future healthcare professionals.

• Algorithm: A set of step-by-step instructions that tells a computer how to solve a problem or complete a task.

• Application programming interface (API): A set of rules and tools that allows different software applications to communicate with each other.

• Artificial intelligence (AI): Computer systems designed to perform tasks that typically require human intelligence, such as visual perception, speech recognition, decision-making, and language translation.

• Artificial general intelligence (AGI): A hypothetical form of AI that would match or exceed human-level cognition across virtually all domains.

• Artificial super intelligence (ASI): A theoretical form of AI that would surpass human intelligence across all domains and continue to self-improve.

• Augmented intelligence: AI systems designed to enhance rather than replace human intelligence and decision-making, particularly relevant in healthcare settings.

• Bard: A large language model developed by Google, that has been rebranded to Gemini.

• Bias: Systematic errors in AI outputs that reflect prejudices present in training data or model design. In healthcare, bias can lead to inequitable care recommendations.

• Chatbot: A computer program designed to simulate conversation with human users, especially over the internet.

• ChatGPT: A conversational interface developed by OpenAI that utilizes GPT models to engage in dialogue with users.

• Claude: An AI assistant developed by Anthropic, known for its attempt at more ethical and controlled responses.

• DALL-E: An AI system developed by OpenAI that creates images from text descriptions.

• Fine-tuning: The process of further training a pre-trained AI model on a specific dataset to adapt it for particular applications, such as medical knowledge.

• Foundation model: Large-scale AI models trained on broad data that can be adapted to many different tasks through fine-tuning or prompting.

• Gemini: Google’s latest large language model, successor to Bard.

• Generative adversarial network (GAN): A type of AI system where two neural networks compete to generate realistic content, often used in medical image synthesis.

• Generative artificial intelligence (gAI): AI systems that can generate new content including text, images, audio, and code, based on patterns learned from training data.

• Generative pre-trained transformer (GPT): A type of large language model architecture developed by OpenAI that learns to predict and generate text through pre-training on vast datasets.

• Hallucination: When an AI model generates information that appears plausible but is factually incorrect or made up. This is particularly concerning in healthcare contexts where accuracy is vital.

• Large language model (LLM): Neural network-based systems trained on vast text datasets to predict and generate human-like text. Examples include GPT-4, Claude, and Llama.

• LLaMA: Meta’s open-source large language model series, designed to facilitate AI research and development.

• Machine learning (ML): A subset of AI where computers learn from data and improve their performance without being explicitly programmed.

• Med-PaLM: Google’s medical-focused version of their PaLM language model, specifically trained on healthcare data.

• Midjourney: An AI system that generates images from text descriptions, known for its artistic and creative outputs.

• Multimodality: The ability of AI systems to work with multiple types of input and output data (text, images, audio, etc.). Increasingly important for clinical applications.

• Natural language processing (NLP): Technology that helps computers understand, interpret, and respond to human language in a useful way.

• OpenChat: An open-source chatbot platform designed to compete with proprietary solutions.

• PaLM: Google’s Pathways Language Model, a large language model preceding Gemini.

• Perplexity: A metric used to evaluate language models’ performance by measuring how well they predict text.

• Prompt: The text input provided to a language model that elicits a response. In healthcare education, crafting effective prompts is key to getting useful outputs.

• Retrieval-augmented generation (RAG): A technique that enhances language model outputs by incorporating information from external knowledge sources, improving accuracy for domain-specific tasks.

• Stable Diffusion: An open-source AI system for generating images from text descriptions.

• Strong AI: Another term for artificial general intelligence (AGI), referring to AI systems that can match human-level cognition.

• Supervised machine learning: A type of machine learning where the algorithm learns from labeled training data.

• Unsupervised machine learning: A type of machine learning where the algorithm finds patterns in unlabeled data.

• Vicuna: An open-source large language model trained on conversational data.

• Weak AI: Also called narrow AI, refers to AI systems designed for specific tasks without general human-like intelligence.