Gen AI 101
This Latin term for "body" is used to refer to the large collection of text data used as input for training large language models.
What is corpus?
This type of generative model learns patterns in text data and can generate responses based on those patterns.
What is a Large Language Model, or LLM?
This refers to systematic and unfair discrimination in GenAI outputs as a result of training data, algorithm design, or human interaction
What is bias?
GitHub recently announced these two new models for use with their auto-complete and chat tools in popular IDEs.
What is GPT 4o and Claude 3.5 Sonnet?
This term is used to describe the ability to leverage text, images, video, and even sound as inputs to interact with models
What is multimodal AI?
This is the practice of human input, oversight, and/or interaction in the decision-making process of an AI system.
What is "human in the loop", or HITL?
Using existing text as source material to augment a language model's knowledge when responding to questions.
What is Retrieval-Augmented Generation, or RAG?
This is what happens when a model produces output that is either not grounded in truth or does not correspond to the facts presented during input
What is hallucination?
Recent reports show that AI-created images have hit this number of images in only 1.5 years. In comparison, human photography is estimated to have taken 149 years to reach this milestone.
What is 15 billion images?
This is considered a next generation of artificial intelligence, decoupled from human intervention.
What is Autonomous AI?
Unlike generative, this form of AI is used in machine learning and designed to predict targets based on input data
What is discriminative AI?
This subfield of AI focuses on interaction between humans and computers through natural language
What is Natural Language Processing, or NLP?
Similar to "shadow IT", this term is used to describe the personal use of AI for business purposes.
What is "shadow AI"?
In a June 2024 report from AWS, this percentage of the web-based text on the internet has been translated with AI.
What is 57%
Designed to perform specific tasks autonomously, this term is given to small computer programs that act on inputs and feedback as part of a broader set of applications.
What is an AI agent?
This refers to data processing that uses interconnected nodes or neurons in a layered structure that resembles the human brain
What is neural network?
This type of machine learning is where an agent learns to make decisions in sequence by executing actions in an environment with some notion of cumulative reward or gains
what is Reinforcement Learning, or RL?
This can happen when the model is not adequately trained, or when the training data becomes overly simplified and biased.
What is model collapse?
Amazon, Alphabet, Meta, and Microsoft are estimated to spend this much in 2024 on datacenter operations to support gen AI. (in billions)
What is $200 billion?
By the end of 2026, estimates are that this percentage of synthetic customer data will be created by generative AI.
What is 76%?
This term is the practice of lowering the precision used to represent model data points in an effort to reduce complexity and perform calculations faster.
What is quantization?
This is the abbreviation for the capability to learn long-term dependencies, often used in generative models
What is Long Short-Term Memory, or LSTM?
The provisional agreement in the EU applies measures to prohibit scraping of images for facial recognition, create categorization systems that could be create discrimination/bias, and using AI for manipulation.
What is the Artificial Intelligence Act?
This type of model is targeting AI to run robots and robotic systems, personalized health interventions, and work towards simulating human behaviors
What is Large Behavior Models, or LBMs?
Forbes recently wrote predictions, including this many robotic humanoids being deployed in the world by 2030.
What is 100,000?