AI Glossary: Understanding Key Terms of Innovation

AI Glossary

Artificial intelligence (AI) is transforming our world, introducing a new vocabulary to describe its mechanisms and impact. Whether you are attending a product meeting, a presentation, or a panel, you will likely hear terms like LLMs, RAG, RLHF, and many others. These terms can confuse even the most seasoned tech experts. This glossary aims to clarify these concepts by providing simplified definitions of the most common AI terms. Whether you are a developer, an investor, or simply curious about tech news, this guide is designed to help you navigate the complex world of AI. We regularly update this glossary to reflect the rapid developments in the field, just like the AI systems it describes.

General Artificial Intelligence (AGI)

General artificial intelligence, often abbreviated as AGI, is a concept that remains vague for many. It generally refers to an AI capable of outperforming the average human in many tasks, if not most. Sam Altman, CEO of OpenAI, describes AGI as the equivalent of an average human that one could hire as a colleague. According to OpenAI's charter, AGI consists of highly autonomous systems that surpass humans in most economically valuable work. Google DeepMind, for its part, defines AGI as an AI that is at least as competent as humans in the majority of cognitive tasks. This diversity of definitions shows that even AI experts do not fully agree on what AGI is.

AI Agent

An AI agent is a tool that uses AI technologies to perform a series of tasks on your behalf, going far beyond the capabilities of a simple chatbot. These agents can handle a variety of tasks such as expense reporting, ticket booking, or even writing and maintaining code. However, this space is rapidly evolving, and the term "AI agent" can have different meanings depending on the context. The infrastructure necessary to fully realize the capabilities of these agents is still under development. The fundamental concept relies on an autonomous system capable of leveraging multiple AI systems to execute complex multi-step tasks.

API Endpoints

API endpoints can be imagined as "hidden buttons" within software that other programs can activate to make it function. Developers use these interfaces to create integrations, allowing, for example, an application to pull data from another or for an AI agent to control third-party services without human intervention. Most smart devices and connected platforms have these endpoints, even if ordinary users never see them. As AI agents gain capabilities, they become increasingly adept at locating and using these endpoints autonomously, thus opening up powerful, sometimes unexpected, possibilities for automation.

Chain of Thought

When a human brain is faced with a simple question, it can often respond without much thought, such as "which animal is bigger, a giraffe or a cat?" However, for more complex problems, it is often necessary to break the question down into intermediate steps. For example, if a farmer has chickens and cows, and together they have 40 heads and 120 legs, it may be necessary to write an equation to find the answer (20 chickens and 20 cows). In the context of AI, chain of thought reasoning for large language models involves breaking a problem down into smaller steps to improve the quality of the final result. Although this generally takes more time, the answer obtained is often more accurate, especially in logical or coding contexts. Reasoning models are developed from traditional language models and optimized for chain of thought through reinforcement learning.

Coding Agent

A coding agent is a more specific concept than an AI agent. It is a program capable of taking actions autonomously, step by step, to achieve a goal in the field of software development. Rather than simply suggesting code for a human to review, a coding agent can write, test, and debug code autonomously, thus managing the iterative and trial-and-error work that typically occupies a developer's day. These agents can operate across entire codebases, spotting bugs, running tests, and applying fixes with minimal human supervision. Imagine a very fast intern who never sleeps and never loses focus, although a human must always review the work.

Computation

The term computation, while multifaceted, generally refers to the essential processing power that enables AI models to function. This type of processing fuels the AI industry, giving it the capability to train and deploy its powerful models. The term is often used as shorthand to refer to the types of hardware that provide this computational power, such as GPUs, CPUs, TPUs, and other forms of infrastructure that form the backbone of the modern AI industry.

Deep Learning

Deep learning is a subset of machine learning that improves autonomously. AI algorithms are designed with a structure of multi-layer artificial neural networks (ANNs), allowing them to make more complex correlations compared to simpler machine learning systems, such as linear models or decision trees. The structure of deep learning algorithms is inspired by the interconnected pathways of neurons in the human brain. Deep learning AI models can identify important features in data themselves, rather than requiring human engineers to define these features. The structure also supports algorithms that can learn from their mistakes and, through a process of repetition and adjustment, improve their own outcomes. However, deep learning systems require many data points to yield good results (millions or more). They also generally take longer to train compared to simpler machine learning algorithms, which tends to increase development costs.

Diffusion

Diffusion is the technology at the heart of many AI models generating art, music, and text. Inspired by physics, diffusion systems "slowly destroy" the structure of data, such as photos or songs, by adding noise until nothing remains. In physics, diffusion is spontaneous and irreversible, like sugar dissolving in coffee that cannot be restored to a cube. However, AI diffusion systems aim to learn a "reverse diffusion" process to restore the destroyed data, thereby acquiring the ability to recover data from noise.

Distillation

Distillation is a technique used to extract knowledge from a large AI model using a "teacher-student" model. Developers send queries to a teacher model and record the outputs. The responses are sometimes compared to a dataset to assess their accuracy. These outputs are then used to train the student model, which is trained to mimic the teacher's behavior. Distillation can be used to create a smaller, more efficient model based on a larger model with minimal distillation loss. This is likely how OpenAI developed GPT-4 Turbo, a faster version of GPT-4.

Fine-Tuning

Fine-tuning refers to the additional training of an AI model to optimize its performance for a task or domain that is more specific than what was previously the focus of its training, typically using new specialized data. Many AI startups take large language models as a starting point to build a commercial product but seek to enhance utility for a target sector or task by complementing previous training cycles with fine-tuning based on their own domain-specific knowledge and expertise.

Generative Adversarial Network (GAN)

A GAN, or generative adversarial network, is a type of machine learning framework that underpins some significant developments in generative AI regarding the production of realistic data, including (but not limited to) deepfake tools. GANs involve the use of a pair of neural networks, one of which relies on its training data to generate an output that is passed to the other model for evaluation. Both models are essentially programmed to try to outdo each other. The generator tries to get its output past the discriminator, while the discriminator works to spot artificially generated data. This structured competition can optimize AI outputs to be more realistic without requiring additional human intervention. While GANs work best for narrower applications, such as producing realistic photos or videos, rather than for general-purpose AI.

Hallucination

Hallucination is the industry's preferred term for AI models that invent information, literally generating incorrect information. Obviously, this is a huge problem for AI quality. Hallucinations produce GenAI outputs that can be misleading and could even pose real-world risks, with potentially dangerous consequences, such as a health query returning harmful medical advice. The issue of AIs fabricating information is considered a consequence of gaps in training data. Hallucinations contribute to a push towards increasingly specialized and/or vertical AI models, that is, domain-specific AIs that require narrower expertise as a means of reducing the likelihood of knowledge gaps and decreasing the risks of misinformation.

Inference

Inference is the process of executing an AI model. It is the act of unleashing a model to make predictions or draw conclusions from data it has already seen. To be clear, inference cannot occur without training; a model must learn patterns in a dataset before it can effectively extrapolate from that training data. Many types of hardware can perform inference, ranging from smartphone processors to powerful GPUs to custom-designed AI accelerators. But not all can execute models equally. Very large models would take ages to make predictions on, say, a laptop compared to a cloud server with high-end AI chips.

Large Language Model (LLM)

Large language models, or LLMs, are the AI models used by popular AI assistants, such as ChatGPT, Claude, Google's Gemini, Meta's Llama, and others.