T

A parameter that controls the randomness and creativity of an AI model's outputs. A temperature of 0 makes the model highly deterministic and predictable (always choosing the most likely next word). Higher temperatures (0.7–1.0) produce more creative, varied, and sometimes unexpected outputs. Most AI tools set temperature automatically, but APIs allow manual control.

AI technology that generates images from text descriptions (prompts). Tools like Midjourney, DALL-E 3, Stable Diffusion, and Ideogram can create photorealistic images, illustrations, logos, and artwork from a written description. This technology has transformed graphic design, marketing, and content creation.

AI technology that converts written text into natural-sounding spoken audio. Modern TTS systems like ElevenLabs, OpenAI TTS, and Google WaveNet produce voices that are nearly indistinguishable from human speech. TTS is used in podcasts, videos, audiobooks, customer service bots, and accessibility tools.

AI technology that generates video clips from text descriptions. Tools like Sora (OpenAI), Runway, and Kling can create realistic video footage, animations, and cinematic sequences from a written prompt. Text-to-video is transforming content creation, marketing, and entertainment production.

The basic unit of text that AI language models process. A token is roughly equivalent to 3–4 characters or about 0.75 words in English. AI models have limits on how many tokens they can process (context window) and charge for API usage based on token count. 'The quick brown fox' is approximately 4 tokens.

The process of breaking text into smaller units (tokens) that an AI model can process. Different tokenizers handle text differently — some split on spaces, others on subword units. Understanding tokenization helps explain why AI models sometimes struggle with unusual spellings, non-English text, or counting characters.

The process of breaking text into smaller units called tokens before feeding it into an AI model. Tokens are typically word fragments — for example, 'unbelievable' might be split into 'un', 'believ', 'able'. Most LLMs process roughly 750 words per 1,000 tokens. Understanding tokenization helps explain why AI models sometimes struggle with character-level tasks like counting letters or rhyming.

The ability of an AI model to use external tools — such as web search, code execution, calculators, or APIs — to complete tasks. Tool use is what transforms a language model from a text generator into an active agent that can interact with the real world and retrieve up-to-date information.

The degree to which a website or content creator is recognized as an expert source on a specific topic by both search engines and AI systems. Building topical authority requires creating comprehensive, accurate, and consistently updated content on a focused subject area. High topical authority is one of the strongest signals for GEO citation.

The dataset used to teach an AI model. For large language models, training data typically consists of billions of text documents from the internet, books, and other sources. The quality and breadth of training data directly determines what an AI model knows and how well it performs.

A technique where a model trained on one task or dataset is adapted for a different but related task. Transfer learning is the reason fine-tuning works: instead of training from scratch, you take a pre-trained foundation model and transfer its learned knowledge to a new domain. It dramatically reduces the data and compute needed to build specialized AI applications.

The neural network architecture that underlies virtually all modern large language models. Introduced by Google in the 2017 paper 'Attention Is All You Need,' the transformer architecture uses a mechanism called self-attention to process sequences of data in parallel, enabling the training of much larger and more capable models than previous architectures.