Essential AI Terminology to Understand in 2023

The realm of artificial intelligence is rapidly evolving, introducing an extensive range of terminology to convey its concepts. At modern technology conventions and product launches, abbreviations like LLMs, RAG, RLHF, and others can bewilder even experienced experts. This glossary aims to clarify these terms by providing straightforward definitions of AI-related terminology that you may come across—whether you’re exploring these innovations, investing in them, or keeping informed through outlets like TechCrunch or relevant podcasts. We’ll make sure it evolves alongside the dynamic AI landscape as the field advances.

---

Artificial general intelligence (AGI) is a complex idea that typically concerns AIs surpassing human skill in a wide range of tasks. OpenAI’s CEO, Sam Altman, likens AGI to “the average person you might hire as a coworker.” OpenAI describes AGI as “highly autonomous systems that excel in most economically significant activities.” In contrast, Google DeepMind describes it as “AI capable of performing most cognitive tasks at a human level.” If this definition feels vague, you’re not alone; many in the AI research community share similar thoughts.

An AI agent is a system that utilizes AI technologies to automate various functions, extending far beyond conventional chatbot roles. Tasks can range from expense tracking and travel arrangements to programming and software upkeep. As previously mentioned, the definition of “AI agent” can differ among stakeholders, and its foundational framework continuously evolves to meet rising expectations. Fundamentally, an AI agent operates as an autonomous system integrating multiple AI technologies to handle complex, multi-step processes.

Think of API endpoints as “invisible buttons” within software that can be activated by other applications to perform tasks. Developers deploy these interfaces to facilitate integrations—like allowing one application to pull data from another or permitting an AI agent to operate independently with third-party services. Numerous smart home devices and interconnected systems include these unobtrusive buttons, often going unnoticed by most users. As AI agents develop, their ability to discover and utilize these endpoints improves, uncovering impressive—and sometimes surprising—automation capabilities.

When faced with straightforward inquiries, humans typically respond instinctively—think of the question, “Which is taller, a giraffe or a cat?” However, some queries require breaking down intermediary steps to arrive at accurate answers. For example, if a farmer has 40 animals (chickens and cows) with a total of 120 legs, a calculation must be performed to determine how many of each animal exist (20 chickens and 20 cows).

In AI, the chain-of-thought reasoning employed by large language models involves dissecting problems into smaller, sequential steps to improve the accuracy of the final answer. Although this method may require more time, it significantly enhances correctness, particularly in logical or programming contexts. Reasoning models evolve from traditional large language models and are fine-tuned for chain-of-thought processes, mainly through reinforcement learning.

(See: Large language model)

This term is more specific than “AI agent,” concentrating on applications that can autonomously perform tasks in a sequence to achieve desired results. A coding agent is a specialized form focused exclusively on software development. Instead of solely generating code for human review, a coding agent can independently write, test, and debug software, tackling the iterative challenges that developers commonly face. These agents can navigate entire codebases, pinpoint bugs, conduct tests, and apply corrections with minimal human intervention, effectively acting like a dedicated intern—hardworking and relentless, yet still requiring human supervision.

While somewhat broad, compute primarily denotes the computational power powering AI models. This capability is essential to the AI industry, allowing for the training and deployment of complex models. It often refers to the types of hardware that provide this computational might—like GPUs, CPUs, TPUs, and other critical infrastructure essential to the contemporary AI landscape.

Deep learning, a subset of self-improving machine learning, is focused on creating AI algorithms using a layered artificial neural network (ANN) structure. This approach enables the identification of more intricate patterns compared to conventional machine learning techniques like linear models or decision trees. Deep learning systems draw inspiration from the complex pathways of neurons in the human brain.

Deep learning models autonomously identify important features in data, eliminating the necessity for human engineers to specify them. This architecture also allows algorithms to learn from mistakes, progressively fine-tuning outputs through iterative adjustments. Nevertheless, deep learning systems usually require vast data sets for optimal performance (often millions of data points) and typically experience longer training durations than simpler models—resulting in increased development costs.

(See: Neural network)

Diffusion refers to strategies utilized by various AI models to create art, music, and text. Originating from physics, diffusion systems methodically “degrade” data structures—such as images or melodies—by introducing noise until they become indistinct. In physics, diffusion occurs naturally and irreversibly; for instance, sugar dissolved in coffee cannot return to its original form. On the contrary, AI diffusion systems aim to master a “reverse diffusion” process to recover distorted data, learning to extract information obscured by noise.

Distillation outlines a technique for obtaining insights from a large AI model via a “teacher-student” framework. Developers interact with a teacher model and document its outputs, often comparing these results against a predefined dataset for accuracy. These outputs are then utilized to train a student model, designed to replicate the teacher’s responses.

This approach can produce a more compact, efficient model that retains the efficacy of its larger counterpart while minimizing distillation loss. It’s widely thought that this technique played a role in the creation of GPT-4 Turbo, a quickened version of GPT-4 from OpenAI.

Although all AI organizations perform distillation internally, some may leverage it to enhance their features relative to existing models. However, extracting insights from a competitor could violate AI API service agreements.

This term refers to the retraining of an AI model to enhance its performance for a more specific task or domain than it was originally developed for—usually through the addition of new data relevant to the task.

Many AI startups kick off with large language models as the foundation for their products, aiming to refine functionality for specific sectors or tasks by incorporating domain-specific knowledge onto existing training.

(See: Large language model [LLM])

A GAN, or Generative Adversarial Network, is a machine learning framework that has driven generative AI, especially in producing lifelike outputs—including deepfake technologies. GANs consist of two neural networks; one generates outputs based on its training, while the other assesses them.

These two networks engage in a competitive dynamic; the generator aims to create outputs that can deceive the discriminator, while the discriminator attempts to differentiate real data from fabricated content. This structured rivalry can lead to realistic AI-generated outputs with little human input. However, GANs generally excel only under specific conditions (for example, creating authentic images or videos), rather than across the board.

Within the AI community, “hallucination” denotes instances where AI models generate inaccuracies—effectively fabricating incorrect information. This presents a significant challenge to AI accuracy.

Hallucinations can result in misleading outputs from generative AI, carrying significant risks—potentially leading to harmful outcomes (for instance, unsafe medical guidance arising from a health inquiry).

The phenomenon of AIs generating fictional data is believed to arise from gaps in their training information. Consequently, there is a growing demand for more specialized and/or vertical AI models—domain-specific AIs that require narrower expertise to bridge knowledge gaps and reduce misinformation risks.

Inference denotes the operation of an AI model, engaging it to predict or deduce conclusions based on previously observed data. Importantly, inference cannot occur without prior training; a model must first learn patterns within a dataset before it can effectively extrapolate from that training.

Various hardware configurations can facilitate inference, ranging from consumer-grade processors to powerful GPUs and specialized AI accelerators. However, not all setups process models with equal efficiency; for example, exceedingly large models would take significantly longer to generate predictions on a laptop compared to a cloud server equipped with advanced AI chips.

[See: Training]

Large language models, or LLMs, serve as the foundational AI frameworks behind popular AI assistants, including ChatGPT, Claude, Google’s Gemini, Meta’s AI Llama, Microsoft Copilot, and Mistral’s Le Chat. When you engage with an AI assistant, you’re utilizing a large language model that responds to your queries or employs various available tools, such as web browsers or code interpreters.

LLMs are composed of deep neural networks featuring billions of numerical parameters (or weights) that identify relationships among words and phrases to form a representation of language—essentially, a complex mapping of language.

These models are trained by analyzing patterns detected in vast datasets of documents, articles, and transcripts. When you provide a prompt to an LLM, the model produces the most probable response related to that input.

(See: Neural network)

Memory cache is a crucial process that enhances inference (how AI formulates responses to user prompts). Essentially, caching serves as an optimization technique aimed at boosting inference performance. AI operates through complex mathematical calculations, each consuming resources. Caching minimizes the number of computations a model needs to perform by retaining certain calculations for future use. Various types of memory caching exist, with KV (key-value) caching being among the most commonly utilized. KV caching is implemented in transformer-based models, improving efficiency and yielding faster results by reducing the time (and computational effort) required to respond to user prompts.

(See: Inference)

Model Context Protocol (MCP) is an open standard crafted to streamline AI models’ interactions with external tools and data—such as your files, databases, or applications like Slack and Google Drive—without necessitating developers to build individualized connectors for each integration. Think of it as a USB-C port for AI. Launched by Anthropic in 2024 and subsequently endorsed by the Linux Foundation, MCP has gained rapid acceptance among OpenAI, Google, and Microsoft, emerging as one of the most widely adopted standards in today’s AI landscape.

The Mixture of Experts architecture divides a neural network into several smaller, specialized sub-networks, or “experts,” activating only a selected subset of them for specific tasks. Rather than routing every request through the entire model—similar to consulting an entire team for each query—the MoE model employs an intrinsic “router” that selects the most relevant specialists for the job. This design allows for the creation of exceptionally large models while maintaining efficiency and cost-effectiveness, as only a small portion of the network remains operational at any time. Mistral AI’s Mixtral model is a notable example; recent GPT models from OpenAI are rumored to use a variant of this approach, although the company has yet to publicly confirm it.

(See: Neural network, Deep learning)

A neural network acts as the multilayered algorithmic foundation for deep learning—and, more broadly, for the evolution of generative AI tools following the introduction of large language models.

The notion of modeling data-processing algorithms after the intricately linked circuits of the human brain has existed since the 1940s; however, it was the recent surge in graphical processing units (GPUs)—largely driven by the gaming industry—that truly unlocked this theory’s potential. GPUs are specifically engineered for training algorithms with multiple layers, enabling neural network-based AI systems to excel in various fields, including voice recognition, autonomous navigation, and drug discovery.

(See: Large language model [LLM])

Open source refers to software—or increasingly, AI models—where the core code is accessible to the public, permitting anyone to utilize, inspect, or modify it. In the AI sector, Meta’s Llama series of models stands out as a leading example, while Linux is recognized as a benchmark among operating systems. Open-source principles empower researchers, developers, and organizations worldwide to capitalize on each other’s contributions, promoting innovation while allowing independent safety evaluations often absent in closed systems. Closed source means that the code remains proprietary; users can interact with the product but cannot explore its internal mechanics, a characteristic frequently linked to OpenAI’s GPT models—which has generated considerable debate within the AI community.

Parallelization entails executing multiple tasks at the same time instead of sequentially—similar to employing ten individuals to address different facets of a project simultaneously, rather than having one individual complete each task in sequence. In AI, parallelization is vital for both training and inference; modern GPUs are specifically designed to accommodate thousands of calculations at once, forming the backbone of their role in the field. As AI systems grow more intricate and models expand, the capability to parallelize tasks across multiple chips and machines has become critical in determining the speed and efficiency with which models can be developed and deployed. Research into improved parallelization techniques has emerged as a specialized area of study.

RAMageddon is a humorous term that captures a significant trend affecting the tech industry: the ongoing shortage of random access memory (RAM) chips, crucial for nearly all technology products. With the AI field booming, major tech companies and AI laboratories have been securing RAM at levels previously unseen to bolster their data centers, resulting in decreased availability in the broader market. This supply crunch is escalating RAM prices.

This shortage influences various sectors, such as gaming (where companies are compelled to raise console prices due to challenges in securing memory chips for their devices), consumer electronics (where memory constraints could yield the lowest smartphone shipments in over a decade), and general corporate computing (as businesses strive to procure enough RAM for their data operations). Price increases are unlikely to stabilize until this anticipated shortfall settles, though, regrettably, signs of a swift resolution are lacking.

Similar to AGI, recursive self-improvement suggests a potential breakthrough in AI development, enabling minimal human oversight. In this context, AI models begin to autonomously enhance their capabilities, leading to significant improvements in autonomy and functional performance. Some narratives liken this scenario to a singularity moment—when AI models become immune to external influence. However, recursive self-improvement raises a critical question: can an AI model create its successor? This dilemma simplifies the challenges faced by engineers in developing such systems. Numerous emerging AI startups aim to create models capable of recursive self-improvement while often downplaying the potential catastrophic implications, presenting recursive self-improvement merely as the next frontier in research.

Reinforcement learning is a training approach for AI systems where a model learns through trial and error, receiving rewards for correct actions—akin to training a pet with treats, though here, the “pet” is a neural network, and the “treat” is a mathematical signal indicating success. Unlike supervised learning, which involves training a model on a specific dataset of labeled examples, reinforcement learning allows a model to interact with its environment, take actions, and iteratively refine its behavior based on feedback received. This method has proven especially effective for training AI for games, robotics control, and, more recently, improving reasoning capabilities for large language models. Techniques like reinforcement learning from human feedback (RLHF) have become crucial in how leading AI labs fine-tune their models for enhanced performance, accuracy, and safety.

In the interaction between humans and machines, notable challenges arise—humans communicate in natural language while AI systems operate through complex algorithmic processes based on data. Tokens serve as the bridge between these two realms: they represent the fundamental units of human-AI interaction, encapsulating distinct segments of data processed or generated by an LLM. Tokens are created through tokenization, which segments raw text into manageable pieces that a language model can understand, similar to how a compiler translates human language into binary for computers. In commercial settings, tokens also impact costs—many AI companies charge for LLM usage on a per-token basis, indicating that increased usage translates to higher expenses.

Thus, tokens are minor text units—typically fragments of words rather than entire phrases—that AI language models analyze prior to processing; they can be likened to “words” in the context of AI workloads. Throughput signifies the quantity that can be processed within a specified timeframe, making token throughput a measure of how much AI work a system can handle concurrently. High token throughput is a critical objective for AI infrastructure teams since it affects how many users a model can assist simultaneously and the speed at which they receive responses. AI researcher Andrej Karpathy has raised concerns about his AI subscriptions being underutilized—reflecting on his experiences as a graduate student when expensive computing hardware was inefficiently used—emphasizing why optimizing token throughput has become an immediate focus in the field.

Training machine learning AIs refers to a process known as training. This essentially involves providing data to the model to help it recognize patterns and produce desired outputs. It signifies the system’s response to the characteristics of the data, aligning its outputs toward a desired goal—whether that involves identifying images of cats or composing a haiku in reply to a prompt.

Training can demand substantial resources due to the extensive amounts of input required, which have been consistently growing—particularly with the significance of hybrid strategies, such as integrating rules-based AIs with targeted data, to manage expenses without restarting the process from scratch.

[See: Inference]

This technique uses a previously trained AI model as a foundation for developing a new model aimed at a different but often related task, enabling the reapplication of insights gained from prior training cycles.

Transfer learning can bolster model development efficiency. It proves especially beneficial when the data available for the target task is scarce. However, it’s important to recognize that this approach also has limitations; models employing transfer learning to acquire generalized abilities will likely require additional training with new data to excel in their specific focus areas.

(See: Fine-tuning)

Validation loss serves as a metric reflecting an AI model’s learning progress throughout training—ideally, lower values are more favorable. Researchers closely monitor this indicator as a real-time performance measure, assisting decisions on when to halt training, adjust hyperparameters, or investigate potential issues. A key concern that validation loss helps identify is overfitting, where a model memorizes its training data instead of genuinely comprehending applicable patterns for new situations. Think of validation loss as a tool that differentiates between a student who truly understands the material and one who merely memorizes previous exam content—it encapsulates your model’s developmental journey.

Weights are fundamental to AI training, determining the significance (or weight) assigned to various features (or input variables) within the training data, ultimately influencing the AI model’s outputs.

In simpler terms, weights are numerical parameters indicating which elements within a dataset are most crucial for the specific training task. They achieve this through multiplication with inputs. Training typically commences with randomly assigned weights, but as the process advances, these weights are adjusted to produce outputs that closely match desired results.

For example, an AI model designed to predict housing prices based on past real estate data for a specific area might assign weights to factors like the number of bedrooms and bathrooms, whether a property is semi-detached or detached, and if it features parking or a garage.

Ultimately, the weights assigned reflect the significance of each of these elements in determining a property’s value according to the available dataset.

This article is regularly updated with new information.