New to AI?

Agentic AI refers to an autonomous system that uses reasoning and tools to independently plan, make decisions, and execute multi-step tasks to achieve a high-level goal with minimal human intervention.

The process of ensuring that an artificial intelligence system’s goals, behaviors, and decisions consistently match human intentions, ethical values, and safety requirements. The most extreme version of AI alignment concern centers around human existential risk (often called X-risk) associated with AI.

A software application designed to simulate human conversation through text or voice by using natural language processing (NLP) to interpret user intent and generate contextually relevant responses.

The multidisciplinary field of study and practice that establishes moral guidelines to ensure artificial intelligence is developed and deployed in ways that protect human rights, mitigate bias, and prioritize transparency and safety.

• Fairness: Ensuring the AI doesn’t discriminate against specific groups due to biased training data.
• Transparency: Being open (transparent) about how a model was trained and what data was used.
• Explainability: The ability for the AI (or its creators) to explain why a specific decision or prediction was made from the model.
• Accountability: Defining who is responsible when an AI system makes a mistake or causes harm.
• Privacy: Protecting user data and ensuring that models don’t leak sensitive personal information.

A context window is the maximum amount of information measured in tokens (see below) that an AI model can “read” and consider at one time when generating a response.

A subset of machine learning (ML) using multi-layered neural networks (inspired by the human brain).

The process of taking a pre-trained AI model and further training it on a smaller, specific dataset to specialize it for a particular industry or task.

A subset of Deep Learning focused on creating new content (text, images, code) rather than just classifying existing data.

A specialized electronic circuit designed to perform millions of mathematical calculations simultaneously through parallel processing.

A phenomenon where an AI generates information that sounds confident and plausible but is factually incorrect or nonsensical.

A type of artificial intelligence trained on vast amounts of text data to understand, summarize, and generate human-like language by predicting the most statistically likely next word in a sequence. Well known implementations of LLMs into chatbots include ChatGPT, Claude, and Gemini.

A mathematical formula that measures the “distance” or error between an AI’s prediction and the actual answer, providing a single numerical score that tells the model how well it is performing on a specific task.

Machine learning is at the core of generative AI. Generative models are trained on large datasets to understand the underlying structures and relationships within the data. Systems “learn” patterns from data rather than following explicit programming.

Models analyze large sets of data, learn from the data, and generate new content that resembles what it studied. The models get better at generating content over time the more they practice, and some models can adapt their creations to fit individual preferences or styles. Algorithms enable the model to explore different combinations of features or styles, allowing for unique or innovative creations. 

A framework where different specialized AI agents (e.g., a “Legal Agent” and a “Finance Agent”) collaborate to solve a problem, often double-checking each other’s work.

This branch of AI focuses on the interaction between computers and human language. Generative AI models like ChatGPT use NLP techniques to generate coherent text based on prompts provided by users. 

An AI framework that improves the accuracy and reliability of large language models by fetching relevant, up-to-date information from external knowledge sources to ground the model’s response. 

A machine learning training method where an AI agent learns to make optimal decisions by interacting with an environment, receiving feedback in the form of rewards for good actions and penalties for bad ones.

The basic unit of text (roughly 0.75 words) that AI models process; these are often the primary metric for usage costs and “context window” limits.

A type of neural network architecture that uses a so called “self-attention” mechanism to process entire sequences of data—like sentences or code—simultaneously, allowing it to understand the relationships and context between all parts of the input regardless of their distance from each other.

Generative models require vast amounts of data for training. The quality and diversity of this training data significantly influence the model’s ability to generate realistic and contextually relevant outputs. 

Many generative AI applications are user-friendly, allowing individuals to input their ideas and receive creative outputs. It’s common for users to guide the creative process through prompts, making the experience collaborative.