Enterprises developing artificial intelligence applications can run into challenges with the complexity of AI. Confusion or ambiguity around artificial intelligence glossary terms such as foundation models, brittleness, and neural networks can derail your enterprise AI development project.
This page provides an artificial intelligence glossary of the most important AI terms enterprise leaders should know.
General Artificial Intelligence Glossary Terms
AI Model
An AI model is a program that is trained on data and uses algorithms to make predictions, decisions, or classifications based on new inputs.
Algorithm
Algorithms are the rules and procedures artificial intelligence models use to process data and perform tasks. They instruct models on how to learn, analyze data, and make decisions.
Machine Learning (ML)
Machine learning is a subset of AI that allows machines to learn from data without being explicitly programmed. The main factors in AI vs machine learning are the scope of knowledge, approach to problem-solving, capabilities, limitations, and level of human intervention required. Machine learning generally has a narrower scope but needs less human intervention to create models and make predictions.
Deep Learning
Deep learning is a subset of machine learning using multiple layers of neural networks. This allows models to learn from larger and more complex data sets.
Neural Networks
Neural networks are inspired by the way the neurological network in the human brain. They are a series of algorithms that use interconnected nodes, or neurons, to recognize patterns and relationships in data and solve complex problems.
Generative AI
Gen AI is a form of AI model that is trained on data to generate new content, such as text, images, or multimedia.
Large Language Models (LLMs)
LLMs are AI models trained on large amounts of text data to understand and generate human language. Generative AI is one application of LLMs.
Natural Language Processing (NLP)
NLP is a collection of AI techniques focusing on understanding and processing human language. This enables machines to interact with text and speech.
Natural Language Understanding (NLU)
A subset and evolution of NLP, natural language understanding focuses on training models to comprehend the meaning and context in human language.
Predictive Analytics
Predictive analytics use AI to analyze data to forecast future trends or outcomes. These analytics help support data-driven decision-making.
Foundation Model
Foundation models are large, pre-trained AI models that can be trained and fine-tuned to perform a variety of tasks. For example, OpenAI’s GPT series, Google’s PaLM 2, and Amazon Titan. Foundation models can be closed-source or open-source.
Agentic AI
Agentic AI refers to the use of AI agents, software programs that can collect data and autonomously perform tasks to achieve predefined goals. Giving AI a greater degree of agency allows it to perform more complex tasks and multi-step workflows.
AI Model Data and Training Terms
Training Data
The dataset used to train an AI model is called training data. This data enables the artificial intelligence model to recognize patterns and make decisions based on new data.
Synthetic Data
Synthetic data is artificially generated data meant to mimic real-world data and serve as training data. Synthetic data can be used when there are limited or insufficient datasets to use for model training.
Data Labeling
Data labeling refers to the process of tagging raw data with labels to provide context and help AI models learn from it. Labeled data makes it easier for models to classify information and recognize patterns.
Supervised Learning
Supervised learning is an ML technique that uses labeled data to train algorithms. This allows the model to follow rules and guidelines set by the data and training methods.
Unsupervised Learning
Unsupervised learning is an ML technique for training algorithms that uses unlabeled data. This allows the ML model more freedom to discover unknown patterns or connections in the data.
Reinforcement Learning
Reinforcement learning is an ML technique that rewards AI agents for getting closer to the model goals. This incentivizes the agent to use trial and error to achieve optimal results.
Bias
Bias is the tendency for AI models to produce skewed, unfair, or discriminatory results due to biased training data or algorithm design.
Brittleness
AI models are considered brittle when they perform well under training conditions but fail in real-world applications due to scenarios or data not accounted for in training.
Hallucinations
Hallucinations are outputs generated by AI models that are incorrect or misleading. Hallucinations are generally caused by insufficient data, biases, brittleness, or incorrect assumptions made by the model.
Explainable AI (XAI)
Explainable AI is the methods and systems designed to provide explanations for an artificial intelligence’s decision-making process. These processes allow for more responsible development by increasing the transparency and trust of the AI model.
Is your organization ready for AI?
AI Cloud Infrastructure Terms
AI Cloud Stack
The AI cloud stack is the interconnected layers, services, and tools used to develop and deploy artificial intelligence applications in the cloud.
Cloud Provider
Cloud providers like AWS, Microsoft Azure, and Google Cloud offer infrastructure and services to develop, deploy, and scale AI models and applications.
Model Hub
Model hubs are repositories for pre-trained AI models that sit between the cloud provider and the foundation model. Model hubs such as Hugging Face allow developers to download, train, and deploy models.
API Integration
Application programming interfaces (APIs) allow AI systems to connect to other software applications. Enterprises can use the APIs of popular AI models such as GPT to run AI applications without developing their own AI model.
Enterprise AI Development Terms
AI Maturity Model
The AI maturity model is a framework enterprises can use to rate their current AI readiness and capabilities.
AI Readiness
AI readiness is a measure of an organization’s readiness for AI implementation. It measures data readiness, people, technology and infrastructure, existing AI projects, and company strategy and vision. Take Gigster’s AI Readiness Assessment today.
Governance
Governance refers to the policies, processes, and guidelines enterprises use to guide the development, deployment, and use of AI within the organization.
Looking for help demystifying the terms in this artificial intelligence glossary and guiding AI development for your enterprise? Gigster can help plan and execute your AI application to help your enterprise get the most from artificial intelligence. Share your project today.
