The following are key terms to understand when considering the use of AI. These terms are also included on an attached PDF file.
Automated Decision-Making: Technology making a decision without human involvement. E.g. an automated loan decision, or an aptitude test that is used to make an automated recruitment decision.
Chatbot: Generative AI that responds to text or speech. It simulates human-like conversations.
Deepfakes: Pictures, videos, and/or audio that have been altered or falsely created using AI. Deepfakes can be used to spread disinformation or misinformation. E.g. a deepfake of a YouTube creator offering his followers free phones.
Generative AI: AI that goes beyond solely analyzing data to create new content like text, images, music, and videos in response to prompts by users. Because it can create new and novel data, and it mimics human creations, it can be riskier (e.g. it can create misleading or false content like deepfakes). E.g. Open AI ChatGPT and Google’s Gemini.
Hallucination: An AI output that appears factual but is incorrect or incoherent. E.g. an incorrect weather prediction, or a chatbot naming a person who set the world record for crossing the English Channel entirely on foot (in 14 hours and 51 minutes).
Intellectual Property (IP): any intellectual creation like literary or artistic works, inventions, images, designs, etc.
Machine Learning AI: AI that uses algorithms to analyze data and learn from it. It provides outputs like decisions, recommendations, and predictions. While it is less risky than generative AI, there are still privacy concerns and the possibility of discrimination and bias in its outputs.
Model Card: A short document with information about an AI model, like how the model should be used, and how it has been evaluated across different cultures, demographics, and/or races.
Prompt: Any data or instruction provided to AI to generate an output. In generative AI, this includes things like instructions that are typed or spoken.