Which AI concepts do you already know? Start with the drag-and-drop exercise and then review the glossary. Many of these terms will come up again later in the e-course, so it is useful to become familiar with them now.
The drag-and-drop exercise you have just performed was generated by AI in 10 seconds
an AI system that can perform human intellectual tasks and learn across various domains, just like a human
Example: does not exist yet, but is considered the ultimate form of AI
a technique that enables machines to mimic human intelligence, such as learning, reasoning, and problem-solving
Example: a chatbot that provides customer service
a design principle that combines AI and human input to achieve better decisions or outcome
Example: a system where AI makes a suggestion, but the final decision is made by a human
a set of steps or instructions that a computer follows to perform a task or solve a problem
Algorithms are the foundation of AI systems.
Example: a search algorithm that finds the most relevant web pages based on keywords
biases in AI models that can arise from incorrect or limited training data, leading to unfair outcomes
Example: a facial recognition system that is less accurate for certain ethnic groups
a term that refers to AI models with mechanisms that are not easily understood or interpreted by humans
Example: a deep learning model that makes predictions, but the logic behind the prediction is not transparent
a type of machine learning that uses deep neural networks to learn complex patterns in large amounts of data
Example: image recognition in self-driving cars
a technology that uses AI and deep learning to create highly realistic fake videos or audio, where faces or voices are synthesized or manipulated to mimic someone else
Example: a video in which a celebrity appears to say or do things they have never said or done
the gap in access to technology and digital skills between different groups, which affects how AI is utilized
Example: older adults who are less familiar with digital technologies and therefore struggle to use AI tools
a proposed regulation by the European Union to govern the use of AI systems with the aim of ensuring the safe and ethical application of AI in Europe
Example: guidelines to restrict high-risk AI applications, such as facial recognition in public spaces
AI systems that make their decisions and actions understandable to humans
Example: an AI system that explains why it has made a particular diagnosis in medicine
AI that can generate new data, such as text, images, or sound, based on patterns from existing data
Example: a chatbot that generates text based on user input
a powerful, large language model developed by OpenAI that is trained to understand, generate, and use human-like text for various language-related tasks
Example: ChatGPT, a specific application of GPT, optimized for conversations and designed to communicate with users in a natural and coherent way
when AI generates incorrect or false information that is not based on its training data
Example: a language model that provides a detailed biography of a person who does not exist, complete with fabricated dates, events, and achievements
an AI approach in which humans train, test, or correct AI systems to keep AI outcomes under control and to adjust errors when necessary
Example: AI-generated medical diagnoses that are reviewed and corrected by a human to ensure they are safe for use
an AI model that understands language and can generate text, such as chatbots or automatic translators
Example: Google Translate
an AI model trained on vast amounts of text to generate human-like text and understand language
Example: the large language model GPT-4, which generates sentences that are logical and contextually appropriate
a branch of AI where systems learn and improve from experience and data without being explicitly programmed
Example: a spam filter that learns which emails are unwanted
AI that combines different types of data, such as text, images, and sound, for a more comprehensive analysis
Example: AI that understands both text and images to perform a more thorough search query
a computational model inspired by the human brain, consisting of layers of neurons that process information
Example: used in speech recognition
a language model with open code and architecture that anyone can use, modify, and improve, promoting transparency and collaboration within the AI community
Example: GPT-2 by OpenAI, which is open-source and available for research and experimentation
the input you provide to an AI model, such as a question or a command, to generate a desired output
Example: ‘Write a short story about an adventurous cat.’
carefully crafting questions or commands to get the best possible output from an AI model
Example: ‘Explain to an 18-year-old how AI can be used to improve a company’s cybersecurity. Limit your response to100 words.’
a technique that allows AI both to retrieve information from a database and to generate new content, often used to provide more accurate answers
Example: a search engine that looks up and summarizes information
a parameter that determines how creative or risky an AI model is when generating output
A higher temperature (e.g. 0.8) results in more variation and creativity, while a lower temperature (e.g. 0.2) produces safer, more predictable results.
Example:
Prompt: ‘Describe the ocean.’
Output at temperature 0.2:
‘The ocean is big and blue.’
Output at temperature 0.8:
‘The ocean shimmers like an endless carpet of moving glass.’
the smallest unit of text that a language model processes, such as words or parts of words
Example: In the sentence ‘AI is cool,’ ‘AI’, ‘is’, and ‘cool’ could each be a token.