Skip to Main Content

GenAI Glossary

AI-dubbing

Think of AI dubbing as a tech-savvy language translator for movies and shows. This technology takes the original dialogue from a video and replaces it with a generated audio track in a different language, making it sound as if the characters are naturally speaking that language.
 

AI-music

AI music is a technology where computers, not humans, are the musicians and composers. AI music uses algorithms and artificial intelligence to create melodies, rhythms, and even entire songs. It's like having a digital composer that can generate new tunes or remix existing ones, all on its own. This tech can mimic various music styles, from classical to pop, making it sound like it was created by human artists.
 

AI-text-to-speech

AI text-to-speech is like having a digital narrator for any text you choose. It's a technology that converts written words into spoken words, almost like a robot reading out loud for you. This tool takes the text from websites, documents, or any digital text and turns it into audio, so you can listen instead of reading. For those who don’t like the sound of their own voice, it's also handy for taking a script or notes you’ve written and having them voiced.
 

AI-voice-cloning

AI voice cloning is a technology that can mimic a person's voice so accurately that it's like having a computerised copycat. By analysing recordings of someone's speech, AI learns to reproduce their way of speaking - their tone, accent, and even emotional inflections. This technology can be used for everything from personalizing virtual assistants to dubbing movies in different voices, making it seem as if the original person is speaking in situations where they're not actually present.
 

AI-voice-generator

AI voice generation is where artificial intelligence generates and synthesizes human-like voices, acting as a digital vocal artist. This technology harnesses advanced algorithms and machine learning techniques to produce a wide array of synthetic voices. It can generate voices of different pitches, accents, and even emotional tones, making them sound convincingly human. Unlike traditional voice recording, AI voice generation can create entirely new voices or mimic existing ones, offering flexibility and variety for uses in narration, virtual assistants, and entertainment. The technology can adapt to various contexts, making it capable of speaking any written text with natural cadence and intonation.
 

Bias

When genAI produces an output that confirms or reinforces existing stereotypes or assumptions, the type of bias is often referred to as "confirmation bias". In an AI or machine learning context it's more accurately "algorithmic bias" or "data bias."

Datasets

Datasets are collections of any type of information - like pictures, text, or sounds - that the AI uses to learn and create new content. These datasets are crucial because they shape what the AI can generate. If a dataset is limited or one-sided, the AI's ability to generate diverse and accurate outputs is similarly restricted. Datasets are the foundational knowledge base from which generative AI draws to perform its creative tasks.
 

Hallucinations

Hallucinations refer to instances where an AI system generates incorrect or nonsensical information. This typically occurs when the AI is faced with data or situations it hasn't been adequately trained on or when there's a flaw in its learning process. These inaccuracies can range from minor errors to completely fabricated or irrelevant outputs, highlighting the importance of comprehensive and accurate training for AI systems.
 

Inaccuracies

genAI outputs can sometimes contain inaccuracies, errors, distortions, fabrications or omissions. In the context of artificial intelligence, inaccuracies are like the mistakes or errors an AI system makes in its outputs or decisions. These inaccuracies can happen for various reasons, such as when the AI is trained with incomplete, outdated, or biased data, or when it encounters situations it hasn't been programmed to handle correctly. These inaccuracies in AI can lead to results that are off the mark, not quite right, or even completely incorrect, emphasizing the need for thorough and diverse training of AI systems to ensure their reliability and accuracy.
 

Large Language Models (LLMs)

Large Language Models (LLMs) are powerful artificial intelligence models designed to comprehend, generate, and engage in human language.The functionality of LLMs can be broken down into three key aspects. Firstly, comprehension: LLMs can read and understand text, just like we do when we read a book or an article. Secondly, the generation they can also create new text, whether it’s writing an essay, composing a poem, or even crafting witty responses. Lastly, engagement LLMs can carry on conversations with people, responding contextually and coherently. The large term is because they're trained on vast amounts of data (like books, articles, and websites) and have expansive neural networks. Imagine feeding them a library’s worth of text!

Machine learning

Machine Learning (ML) is a subfield of artificial intelligence that focuses on developing algorithms and models capable of enabling computer systems to learn and improve from experience without explicit programming. In traditional programming, developers provide explicit instructions for a computer to perform a task, but in machine learning, the system learns patterns and insights from data to make predictions or decisions without being explicitly programmed for each specific scenario. The learning process involves exposing the system to large amounts of data, allowing it to identify patterns, relationships, and trends. Machine Learning is categorised into supervised learning, unsupervised learning, and reinforcement learning, each with its own methodologies and applications.

Malicious purpose

Malicious purpose in Gen AI involves intentional use of artificial intelligence to create harmful content, going beyond unintentional errors. This includes generating deceptive information, manipulating data, or crafting content with the explicit aim of causing damage.
 

Neural networks

Neural networks are like the brains of AI systems. They're inspired by how our brains work. Imagine a bunch of interconnected nodes or artificial neurons that team up to process information. These nodes are organised into layers: one for taking in data, one or more hidden layers to process that data, and a final layer to produce the output. Each connection between these nodes has a weight, sort of like the strength of a friendship. During training, the network learns from examples, adjusting these weights to get better at making predictions. It's like learning from mistakes and getting better at a task over time.

Outputs

Outputs in Gen AI are the content or answers the computer gives based on our questions.  They are the end results of the AI's 'thinking' process. These outputs can vary widely in form, such as text, images, or decisions, depending on the AI's design and the task at hand.
 

Prompts

Prompts (or inputs) in Gen AI are the questions we ask to get specific content or answers from the computer. Think of them like questions or instructions given to an AI system, guiding what it should focus on or generate. They are the starting points that set the AI's task in motion. The clarity and precision of these prompts heavily influence the AI's output quality.
 

Prompt engineering

Prompt engineering for Gen AI is designing the right questions to get the best content or answers from an artificial intelligence. It is the skillful crafting of questions or instructions to guide an AI system's responses or creations. This technique is essential for obtaining precise and relevant results from AI, as the way a prompt is structured can significantly influence the AI's output.
 

Training data

In the context of genAI, "training data" refers to the dataset used to teach a generative artificial intelligence system. This data can include text, images, sounds, or any other relevant information that the AI uses to learn patterns, styles, and structures. The quality and diversity of the training data directly influence the AI's ability to generate accurate and nuanced outputs. Essentially, it forms the foundational knowledge base from which the AI draws to create new content.