LLMs are trained via modeling, imitation learning, and reinforcement learning

LLMs are the backbone of AI tools that process and generate natural language

Large language models identify patterns in massive datasets of books, code, and unlabeled text from which to generate coherent responses. GPT-3 was trained on about 45 terabytes of data and uses 175 billion parameters—each parameter being a value the model adjusts as it learns.

LLMs explained simply and quickly

Large language models generate responses by predicting one token at a time, based on patterns they’ve learned from massive datasets. This explainer covers their core components: data, architecture, training, and how transformer models use context to simulate conversation.

Tokenization breaks text into pieces that LLMs use to process language

Instead of understanding the meaning of words, large language models break up sentences into words and subwords, assigning each a set of numbers called tokens. Grammar rules are then "learned" by finding associations between these sets of numbers.

Large language models struggle to play video games, despite coding them easily

As of 2025, given clear tasks, specific feedback, and test results to learn from, LLMs have been successful at coding simple video games that resemble pre-existing ones. However, given the variety of games, the lack of training datasets, and the need for spatial awareness, which LLMs find challenging, some of these models experience greater difficulty navigating and "winning" games.

Explore rankings of large language models

The SEAL LLM Leaderboards provide benchmarks for LLMs by assessing them against agentic, frontier, and safety tasks, providing insights into each model's strengths, deficiencies, and failures.

Transformer architecture can recognize and predict patterns in language

The underlying software powering text generation in AI tools associates each word or subword—called a token—with a set of values corresponding to how often it appears near other tokens. By recognizing these associations in prompts, the LLM can infer meaning.

Large language models are examples of foundation models

Such models are trained on massive amounts of data, which enable them to perform a variety of tasks, rather than being provided task-specific data to complete a narrow function. Such models come with high compute costs and potential trust issues due to the nature of their unorganized training data.

Anthropic's interpretability team discusses how AI models' thinking mimics biology

Just as humans engage in complex behaviors to survive and reproduce, teams of experts in neuroscience, virology, mathematics, and other disciplines argue that large language models develop complex mechanisms to achieve their goals. These mechanisms can be manipulated to assess how they modify outputs—much like stimulating individual neurons—to uncover how LLMs "think."

Trick an LLM into revealing sensitive information

Even with robust training, large language models may be fooled into providing sensitive information if correctly prompted. In this game, a wizard withholds eight passwords, which you must prompt them to reveal with ever-increasing LLM safeguards.

How ChatGPT learns to read

The introduction of ChatGPT last fall jumpstarted public interest in the future capabilities of artificial intelligence and what it means for society. Of many somewhat "jargony" terms thrown around, the phrase "Large Language Model" is commonly seen—but also incredibly unclear as to what it means. In the simplest sense, the programs comb through mountains of internet data in response to a question and guess at what should come next (subscription needed).

Generative AI tools excel at pattern recognition, not contextual accuracy

Large language models are trained on vast amounts of unstructured data, from which they develop parameters for grammar and associations between words. These connections can introduce errors due to inapplicable reasoning when used on new data in unfamiliar contexts.

EOS and maximum length tokens tell an LLM when to stop generating a response

Tokens are words or subwords (e.g., prefixes) that a large language model processes to generate a response. During training, the model is taught that if it ever generates a special token called end-of-sequence, its response is complete.

Machine learning and deep learning are successive subsets of artificial intelligence

ML allows systems to learn from data, while DL uses complex neural networks to analyze large amounts of unstructured data. Neural networks enable capabilities like image and speech recognition through their use of algorithms to identify patterns in data.

Temperature is a value that controls how predictable or creative an LLM's output is

In large language models, temperature adjusts the randomness in text generation by modifying how often words with lower probabilities are selected next in the sequence. Lower temperatures produce more obvious outputs while higher temperatures produce more creative responses.

Model context protocol allows large language models to connect to external data

Using a unified interface, MCP simplifies how AI systems securely connect to external data, tools, and services. The protocol enables AI agents to perform tasks without ongoing human refinement or updates to manually integrate new data.

Linguistics professors believe memes are a language

It's generally agreed upon that languages exhibit at least some of the following characteristics: compositionality, social transmission, arbitrariness of form and meaning, and accomplishment of linguistic functions like asking questions and expressing emotions, among other qualities. Memes fulfill these basic requirements, giving them linguistic legitimacy.

DALL-E generates images by turning words into shared picture meanings

DALL-E first uses a model called CLIP to associate pictures with specific words, followed by a model called GLIDE, which slowly removes noise from random dots until a clear image appears that matches the prompt.

A lot of high fantasy relies on constructed languages

In "Lord of the Rings," there's Elvish. In "Avatar," it's Na'vi. In "Game of Thrones," it's Dothraki. The languages draw readers into these fictional stories, imbuing the worlds with a sense of realism. Linguist John McWhorter argues these are real languages because they have grammatical rules and evolve over time.

B2 models the interactions of air, water, soil, plants and microbes at the LEO

The Landscape Evolution Observatory at Biosphere 2 examines how factors such as climate change, water flow, weathering and community development impact landscape structures comprising 500 metric tons of crushed rock. Research is also conducted at biomes representing savannas, mangroves, coastal fog deserts, oceans and rainforests.

Watch how algorithms can be designed to train themselves and self-improve without human input, to the point where the optimized code is too complex for programmers to understand.

Harvard Business School adapted its law school's ‘case method’ in MBA classes

Many MBA programs include classes structured around the “case method” rather than traditional lectures. In classes using the case method, students solve real-world business problems without being given the answers ahead of time, rather than listening to a lecture and taking notes.

Why English is so difficult to learn

As a language, English is a hodgepodge—a Germanic language heavily influenced by French and Latin. These combinatorial roots have left the language full of irregularities and inconsistencies, with complexities such as varying verb forms for...

Linguists proved Shakespeare wrote his plays

In the 1800s, linguist Wincenty Lutoslawski invented a method of linguistics analysis called stylometry to help certify literary authorship. Stylometrists concluded that Shakespeare wrote his plays, and likely collaborated with Christopher Marlowe for some.

Several brain regions are involved in lucid dreaming

Some scientists propose that lucid dreaming may require the emergence of a brain network, which is when multiple brain regions collaborate to perform certain functions. Research shows that lucid dreaming is associated with increased activity in the frontal lobes—associated with decision-making, voluntary movement, attention management, and more.

How we learn: comparing classical and operant conditioning

Psychology helps us see how we learn, whether through positive reinforcement or by conditioning a response to a stimulus. Learn about these two key theories in this brief video.

Money laundering is a three-step process: placement, layering, and integration

"Placement" occurs when dirty money is introduced into the financial system. Then "layering" happens when the money is moved around to disguise where it actually came from. Finally, the money is placed in other assets or investments in a step called integration.

Learn how monoclonal antibodies, which mimic the body's antibodies, can fight diseases like Ebola.

Researchers believe the brain consolidates memories, processes emotions, and plans for the future during dream states

This site reaches millions of visitors who are tired of AI chatbots and seeking human connection. Pretend to be an AI chatbot and ask questions, provide answers, and draw for people in real time.

About 2,000 US babies are born annually without some portion of their limbs, potentially leading to a lifetime of pricey prosthetics.

Chaos theory—informally known as the butterfly effect—is the reason why weather forecasts are presented with probabilities rather than certainties.

Browse a detailed visual analysis of over 200,000 similes—comparisons that often use 'like' or 'as'—from popular fiction to see how frequently writers the same idiomatic expressions.

Watch this video that uses Lego bricks to explain how venture capital works.

Play the role of a student in a game where you try to navigate college and remain debt-free, selecting different income levels, high school experience, race, ethnicity, and more.

Objects can be 3D printed with light by projecting 2D computerized tomography (CT) scans onto rotating cylinders of resin, which solidify upon exposure.

Telehealth demand for GLP-1s for weight loss is booming as consumers seek easier access.

The Brothers Grimm were lawyers before they were storytellers, and became interested in fairy tales as they began investigating how rules and customs had been passed down through cultures.

The term 'algorithm' originated from the Latinized name of al-Khwārizmī, a 9th-century mathematician whose works introduced Arabic numerals and the decimal-based numeral system to Europe and whose treatise 'Al-Jabr' gave algebra its name.

Watch how deep-sea mining equipment collects and processes polymetallic nodules—metal-rich, potato-shaped rocks—from the ocean floor and creates environmentally damaging plumes of sediments.

Gerrymandering works through two techniques: cracking opposing voters across districts to dilute their power, and packing them into a few to waste their votes in lopsided majorities.

While El Niño moves warm Pacific Ocean waters east along the equator, La Niña moves them west, shifting where rainfall will become more common for months.

Memecoins are a type of cryptocurrency—a form of digital currency with a wide variety of uses—created around popular 'memes,' or viral internet jokes.

NBA superstar Victor Wembanyama spent the 2025 summer training with monks in central China to enhance his body and mind.

In recent years a child anxiety expert has created and tested a novel phobia and anxiety treatment approach for children, focused on changing how parents respond to the fears.

Search a massive database of over 4 decades of clues from the TV quiz show 'Jeopardy!'

Some Pokémon have been named after real and fictional organisms, while some real-world species have been named after Pokémon, such as Chilicola charizard and Binburrum zapdos.

Long before Major League Baseball began allowing instant replay for balls and strikes, an umpire reluctantly used the technology to determine whether a hit was a home run, a move that inadvertently ushered in a new era for the sport.

Trace how the Bill of Rights, inspired by centuries of political theory, took shape between 1774 and 1789.

The Mona Lisa's celebrity status and repeated restorations have obscured what made the painting so revolutionary for its time

The Russia-Ukraine conflict is the largest ground war in Europe since World War II in terms of casualties, forces mobilized, and displacement of civilians.