Heard AI Terms? Stop Nodding, Start Understanding.

Artificial intelligence is rapidly transforming the world, simultaneously coining an entirely new vocabulary to articulate its mechanisms. Even a brief engagement with AI topics quickly introduces terms like LLMs, RAG, RLHF, and numerous others, which can leave even highly knowledgeable tech professionals feeling uncertain. This glossary aims to address that challenge. We commit to regular updates as the field progresses, positioning it as a dynamic resource, much like the AI systems it describes.

Artificial general intelligence, or AGI, remains an elusive concept. However, it generally refers to AI systems that possess capabilities superior to an average human across many, if not most, tasks. OpenAI CEO Sam Altman once characterized AGI as the "equivalent of a median human that you could hire as a co-worker." Conversely, OpenAI’s charter defines AGI as "highly autonomous systems that outperform humans at most economically valuable work." Google DeepMind’s perspective diverges slightly, viewing AGI as "AI that’s at least as capable as humans at most cognitive tasks." Feeling confused? Rest assured, even experts at the forefront of AI research share this sentiment.

An AI agent denotes a tool that leverages AI technologies to execute a series of tasks on your behalf—going beyond the scope of a more basic AI chatbot—such as filing expenses, booking tickets or restaurant tables, or even writing and maintaining code. However, as previously explained, this emerging domain involves numerous evolving components, meaning the term "AI agent" may hold different interpretations for different individuals. Infrastructure is also still under development to fully realize its envisioned capabilities. Nevertheless, the fundamental concept implies an autonomous system capable of drawing upon multiple AI systems to accomplish multi-step tasks.

Consider API endpoints as "buttons" on the backend of a software application that other programs can "press" to invoke specific functions. Developers utilize these interfaces to construct integrations—for instance, enabling one application to retrieve data from another, or allowing an AI agent to directly control third-party services without requiring manual human operation of each interface. Most smart home devices and connected platforms feature these underlying buttons, even if ordinary users never see or interact with them. As AI agents become more sophisticated, they are increasingly able to autonomously discover and employ these endpoints, unlocking powerful—and at times unexpected—possibilities for automation.

Presented with a straightforward question, a human brain can typically provide an answer without extensive deliberation—such as identifying which animal is taller between a giraffe and a cat. Yet, in many scenarios, a pen and paper are often necessary to arrive at the correct solution due to intermediary steps. For instance, if a farmer possesses chickens and cows that collectively have 40 heads and 120 legs, one might need to formulate a simple equation to deduce the answer (20 chickens and 20 cows).

Within an AI context, chain-of-thought reasoning for large language models involves dissecting a problem into smaller, intermediate steps to enhance the quality of the final outcome. While this approach typically requires more time to generate an answer, the result is more likely to be accurate, especially in logic or coding scenarios. Reasoning models are developed from traditional large language models and are optimized for chain-of-thought thinking through reinforcement learning.

(See: Large language model)

This concept is more specific than a general "AI agent," referring to a program capable of taking autonomous, step-by-step actions to achieve a goal. A coding agent is a specialized version applied to software development. Rather than merely suggesting code for a human to review and implement, a coding agent can independently write, test, and debug code, handling the iterative, trial-and-error tasks that typically consume a developer’s workday. These agents can operate across entire codebases, identifying bugs, running tests, and pushing fixes with minimal human oversight. Envision it as hiring an exceptionally fast intern who never sleeps and maintains unwavering focus—though, as with any intern, human review of the work remains necessary.

Although a somewhat multifaceted term, "compute" generally refers to the essential computational power that enables AI models to operate. This processing capability fuels the AI industry, providing the means to train and deploy its powerful models. The term often serves as shorthand for the types of hardware that deliver this computational power—such as GPUs, CPUs, TPUs, and other forms of infrastructure that constitute the bedrock of the modern AI industry.

Deep learning is a subset of self-improving machine learning in which AI algorithms are designed with a multi-layered, artificial neural network (ANN) structure. This architecture allows them to identify more complex correlations compared to simpler machine learning-based systems, such as linear models or decision trees. The structure of deep learning algorithms draws inspiration from the intricately interconnected pathways of neurons in the human brain.

Deep learning AI models are capable of identifying important characteristics within data autonomously, rather than requiring human engineers to define these features. This structure also supports algorithms that can learn from errors and, through a process of repetition and adjustment, refine their own outputs. However, deep learning systems necessitate a large volume of data points (millions or more) to yield effective results. They also typically require longer training times compared to simpler machine learning algorithms, leading to higher development costs.

Diffusion is the core technology behind many AI models that generate art, music, and text. Inspired by principles of physics, diffusion systems gradually "destroy" the structure of data—for instance, photos or songs—by incrementally adding noise until nothing remains. In physics, diffusion is spontaneous and irreversible—sugar diffused in coffee cannot be restored to its cube form. However, diffusion systems in AI aim to learn a form of "reverse diffusion" process to restore the destroyed data, thereby gaining the ability to recover the data from noise.

Distillation is a technique utilized to extract knowledge from a large AI model through a ‘teacher-student’ model approach. Developers send requests to a teacher model and record its outputs. These answers are sometimes compared against a dataset to assess their accuracy. The recorded outputs are then used to train the student model, which is optimized to approximate the teacher’s behavior.

Distillation can be employed to create a smaller, more efficient model based on a larger model with minimal distillation loss. This method likely contributed to OpenAI’s development of GPT-4 Turbo, a faster version of GPT-4.

While all AI companies use distillation internally, it may also have been used by some AI companies to catch up with frontier models. Distillation from a competitor’s model typically violates the terms of service of AI APIs and chat assistants.

This refers to the further training of an AI model to optimize its performance for a more specific task or domain than was initially emphasized during its training—typically by feeding in new, specialized (i.e., task-oriented) data.

Many AI startups are adopting large language models as a foundation for building commercial products, but they are striving to amplify utility for a target sector or task by supplementing earlier training cycles with fine-tuning based on their proprietary domain-specific knowledge and expertise.

(See: Large language model [LLM])

A GAN, or Generative Adversarial Network, is a type of machine learning framework that underpins significant advancements in generative AI for producing realistic data—including, but not exclusively, deepfake tools. GANs involve the use of a pair of neural networks: one, the 'generator,' creates an output based on its training data, which is then passed to the other model, the 'discriminator,' for evaluation.

These two models are essentially programmed to compete against each other. The generator attempts to produce an output convincing enough to deceive the discriminator, while the discriminator works to identify artificially generated data. This structured competition can optimize AI outputs to be more realistic without the need for additional human intervention. Though GANs perform best for narrower applications (such as generating realistic photos or videos), rather than for general-purpose AI.

Hallucination is the AI industry’s preferred term for instances where AI models fabricate information—literally generating incorrect data. This obviously poses a significant challenge to AI quality.

Hallucinations produce GenAI outputs that can be misleading and potentially lead to real-life risks—with dangerous consequences (consider a health query that yields harmful medical advice).

The problem of AIs fabricating information is thought to arise from gaps in their training data. Hallucinations are driving a push toward increasingly specialized and/or vertical AI models—i.e., domain-specific AIs requiring narrower expertise—as a strategy to reduce the likelihood of knowledge gaps and mitigate disinformation risks.

Inference is the process of executing an AI model. It involves deploying a model to make predictions or draw conclusions from previously encountered data. To be clear, inference cannot occur without prior training; a model must first learn patterns within a dataset before it can effectively extrapolate from this training data.

Many types of hardware can perform inference, ranging from smartphone processors to powerful GPUs and custom-designed AI accelerators. However, not all of them can run models with equal efficiency. Very large models would take considerably longer to generate predictions on, for instance, a laptop compared to a cloud server equipped with high-end AI chips.

Large language models, or LLMs, are the AI models utilized by popular AI assistants such as ChatGPT, Claude, Google’s Gemini, Meta’s AI Llama, Microsoft Copilot, or Mistral’s Le Chat. When you interact with an AI assistant, you are engaging with a large language model that processes your request directly or with the aid of various available tools, such as web browsing or code interpreters.

LLMs are deep neural networks composed of billions of numerical parameters (or weights, see below) that learn the relationships between words and phrases, thereby creating a multi-dimensional representation of language. These models are developed by encoding the patterns they discover in billions of books, articles, and transcripts. When you provide a prompt to an LLM, the model generates the most probable pattern that fits the prompt.

Memory cache refers to an important process that enhances inference (the process by which AI works to generate a response to a user’s query). In essence, caching is an optimization technique designed to make inference more efficient. AI operations are inherently driven by intensive mathematical calculations, and each time these calculations are performed, they consume more power. Caching is designed to reduce the number of calculations a model might need to run by saving particular calculations for future user queries and operations. There are different kinds of memory caching, although one of the more well-known is KV (or key-value) caching. KV caching functions in transformer-based models, increasing efficiency and driving faster results by reducing the amount of time (and algorithmic labor) it takes to generate answers to user questions.

A neural network refers to the multi-layered algorithmic structure that underpins deep learning—and, more broadly, the entire boom in generative AI tools following the emergence of large language models.

Although the concept of drawing inspiration from the densely interconnected pathways of the human brain as a design structure for data processing algorithms dates back to the 1940s, it was the much more recent rise of graphical processing hardware (GPUs)—via the video game industry—that truly unlocked the power of this theory. These chips proved exceptionally well-suited to training algorithms with many more layers than was possible in earlier epochs—enabling neural network-based AI systems to achieve far better performance across many domains, including voice recognition, autonomous navigation, and drug discovery.

(See: Large language model [LLM])

Open source refers to software—or, increasingly, AI models—where the underlying code is made publicly available for anyone to use, inspect, or modify. In the AI world, Meta’s Llama family of models is a prominent example; Linux serves as the famous historical parallel in operating systems. Open source approaches empower researchers, developers, and companies globally to build upon one another’s work, accelerating progress and enabling independent safety audits that closed systems cannot easily provide. Closed source means the code is private—users can utilize the product but not examine its internal workings, as is the case with OpenAI’s GPT models—a distinction that has become one of the defining debates in the AI industry.

Parallelization means executing many tasks simultaneously instead of sequentially—much like having 10 employees working on different parts of a project concurrently instead of one employee doing everything in order. In AI, parallelization is fundamental to both training and inference: modern GPUs are specifically designed to perform thousands of calculations in parallel, which is a significant reason they became the hardware backbone of the industry. As AI systems grow more complex and models grow larger, the ability to parallelize work across many chips and many machines has become one of the most important factors in determining how quickly and cost-effectively models can be built and deployed. Research into better parallelization strategies is now a field of study in its own right.

"RAMageddon" is the fun new term for a not-so-fun trend sweeping the tech industry: an ever-increasing shortage of random access memory, or RAM chips, which power virtually all the tech products we use daily. As the AI industry has blossomed, the biggest tech companies and AI labs—all vying to have the most powerful and efficient AI—are acquiring so much RAM to power their data centers that there is not much left for the rest of us. This supply bottleneck means that what remains is becoming progressively more expensive.

This impacts industries such as gaming (where major companies have had to raise prices on consoles because it’s harder to find memory chips for their devices), consumer electronics (where memory shortages could cause the biggest dip in smartphone shipments in more than a decade), and general enterprise computing (because those companies can’t get enough RAM for their own data centers). The surge in prices is only expected to cease after the dreaded shortage ends, but unfortunately, there’s not really much of a sign that’s going to happen anytime soon.

Reinforcement learning is a method of training AI where a system learns by trying things and receiving rewards for correct answers—akin to training your beloved pet with treats, except the "pet" in this scenario is a neural network and the "treat" is a mathematical signal indicating success. Unlike supervised learning, ...