AI Glossary

Let's kick-off with the basics: What is...

• Algorithm - provides the instructions. A step-by-step process for the computer to follow as it analyzes the data and learns from it. It's like a recipe that tells you how to transform the ingredients (data) into the final dish (model)

• Model - Imagine a model as the end product of your machine learning project. It's a mathematical representation of the data you've analyzed. This model can be used to make predictions on new, unseen data. Read more Machine Learning...

Algorithm + Test Data = Model (creation) → Model + Unseen Data = Prediction (repeat and refine)

• AI,  Artificial Intelligence - many argue that the true meaning of this term infers a level of humanistic assimilation i.e. autonomy that we have not reached, yet. And so we are currently still in the Machine Learning phase. However the term is now being applied to anything ‘smart’

• Explainable AI, XAI - is a collection of techniques and processes that help people understand the reasoning behind a machine learning algorithm's output. XAI can help improve and debug models, meet regulatory requirements, and increase trust in AI models. Read more...

• AI Regulation and Compliance - A set of governance standards for Providers and Deployers of AI Systems, aiming to offer users assurance when engaging with AI applications. For example : EU AI Act (Europe) ; ISO/IEC 42001 (International) ; UK Pro-Innovation AI Framework and California's SB-1047 . Read more...

• ML, Machine Learning - A branch of AI which could be considered the precursor to current Generative-AI, is the science of developing algorithms and statistical models that computer systems use to perform complex tasks without explicit instructions. More associated with finding commonalities and patterns within large databases of either categorised or uncategorised data - like determining weather patterns.

• Deterministic and Heuristic in Machine Learning - Deterministic Models in Machine Learning typically refer to models where, given a specific input, the output is always the same.  Heuristic Models, on the other hand, often involve elements of randomness or approximation, critical elements of AGI. Read more...

• Predictive Algorithms - These are the mathematical formulas and statistical techniques that form the core of machine learning. They analyze data, identify patterns, and learn relationships between variables. Imagine them as the building blocks.

• Predictive Models - This is the final output you get after training the algorithms on data. It's a representation of the learned patterns and can be used to make predictions on new data. Think of it as the finished structure built from the algorithms.

• Inference - the process by which a trained AI model applies its learned knowledge to make predictions or decisions about new, unseen data. It's essentially the model's moment of truth, where it demonstrates its ability to generalize and reason based on the information it has been exposed to during its training phase. Read more about 'Reasoning'

• at inference time - signifies the specific moment when a trained model is utilized to make predictions or classifications on fresh, previously unseen data. It essentially refers to the phase where the model is actively employed to generate outputs based on the input it receives.

• AGI, Artificial General Intelligence - is basically machines that can think and learn like humans, able to tackle many kinds of problems and tasks. Many consider this capability much closer than we think. Read more...

• Generative AI - Refers to any AI system capable of creating new content, which can be text, images, music, code, or other forms.

• GPAI, General-Purpose AI - is defined as an AI model that exhibits significant generality and can competently perform a wide range of tasks. This type of AI model is trained with a large amount of data using self-supervision at scale. Read how EU AIA differentiates GPAI from traditional software systems and machine learning...

• AI Slop - A term used to describe low-quality, AI-generated content that's produced in large quantities and prioritizes quantity over quality ...flooding our Facebook feeds! Considered widely the successor to SPAM, it can be image, text, audio or video.

• AI Agents - Are AI’s that understand and apply tasks automatically / autonomously. An AI Agent may be a bot that is setup to accept simple instructions / task requests and automatically ‘find’ and apply from various other Ai tools a solution to improve itself and/OR how it performs whatever you have set it up to do. Read more...

• MAS, Multi-Agent System  - is a system composed of multiple interacting intelligent agents. Agents are autonomous entities that can perceive their environment, make decisions, and act upon them to achieve specific goals. MAS are widely used in distributed and complex problem-solving scenarios. Read more...

• Reasoning Frameworks  - provide a structured approach for AI agents to process information, make decisions, and generate responses - as part of the Orchestration loop (see AI Agents). They act as a blueprint for the agent's cognitive architecture, guiding its thought process and actions. Popular and notable ReAct (Reason and Act), Chain-of-Thoughts (CoT), Tree-of-Thought (ToT) and Reflexion. Read more...

• Foundation Model - powerful AI model trained on a massive amount of data to be adaptable for various tasks. Read more...

• Frontier Model - A particularly advanced and powerful type of Foundation Model. All Frontier Models are Foundation Models, but not all Foundation Models are Frontier Models. Read more...

• LLM, Large Language Model (for eg. ChatGPT) is a type of artificial intelligence that can understand and generate human language, trained on massive amounts of text data.

• SLM, Small Language Model perform smaller tasks and require less computing power. Read more...

• Transformer Models -  If next word suggestion is driven by neural network responses with the most likely word based purely on mathematical probability (Machine Learning), adding in a transformer model that has layers of attention steps built into it, will guide the responses to a question and or a statement to have more context and meaning.

• GPT, Generative Pre-trained Transformer - It's a type of artificial intelligence (AI) focused on understanding and generating human language. Essentially, it's a powerful language model that can be used for various tasks like creating different creative text formats or predicting what you might type next. Read more about  Vector Embedding and Transformers...

• Multimodal LLM - a large language model that can understand and generate information from multiple sources, like text, images, and even audio, not just text alone. Read more about Multimodal capability...

• VLM, Vision Language Model - This is a specific type of multimodal model that focuses on understanding the relationship between visual data (images) and textual data (language). Read more...

• MMLU, Massive Multitask Language Understanding - a new benchmark designed to measure knowledge acquired during pretraining by evaluating models exclusively in zero-shot and few-shot settings.

• NLP, Natural Language Processors (of which an LLM is a type of NLP) - Natural Language Processing, is a field of computer science concerned with giving computers the ability to understand and work with human language. It's basically like teaching computers to speak our language.

• GLUE, General Language Understanding Evaluation and SuperGLUE are widely used for evaluating the natural language understanding (NLU) capabilities of AI models. Along with MMLU (Massive Multitask Language Understanding) Read more about Benchmarks...

• Prompt - In the context of LLM, a prompt acts as a set of instructions and information that guides us towards generating the desired response. It essentially tells us what to do and how to do it. Imagine it as a map that steers us in the right direction. Read more...

• RAG,  Retrieval-Augmented Generation - a technique that enhances the accuracy and factual grounding of generative models by incorporating external information retrieval, by breaking down documents you provide into segmented token data chunks. Read more...

• TAG, (Table-Augmented Generation) is an AI technique that enhances generative models by enabling them to directly query structured data tables, such as spreadsheets or databases, to retrieve relevant information. This contrasts with RAG (Retrieval-Augmented Generation), where the model retrieves information from unstructured or semi-structured text sources and then uses this information to generate responses. Below is a table summarizing key differences between TAG and RAG.

• Synthetic Data - is essentially artificial information created by computers, rather than collected from real-world events. Read more...

• Tokens - In the context of AI, particularly in multimodal large language models (LLMs), a token refers to the basic unit of information used for processing and representation. Important : tokens are also the measurement of cost associated with the training, deployment and usage of your models. Read more about Tokenization...

• VectorDB - A Vector Database is a specialized database designed to store and retrieve information based on similarity. Unlike traditional databases that rely on exact matches, vector databases store data as mathematical representations called vectors. Read more...

• Open Source - where the underlying code is freely available for anyone to access, tinker with, and distribute. This openness breeds collaboration and fosters a community effort around AI development. Read more...

← Knowledge