Large language models are examples of foundation models

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.

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.

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

Training large language models begins with pretraining the model to predict the next word in a sequence based on finding patterns in massive amounts of text. Patterns are then fine-tuned to model human-written dialogue and to align responses with user preferences.

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.

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).

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.

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.

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."

Unlike the models made by large AI companies, open-weight models can have their safeguards removed or weakened, a process that has become increasingly easy and may expose users to harmful information.

This removal is made possible by developers making the parameters behind their models publicly known. By adjusting these parameters, users can modify how the model processes and provides information. Many of these models can be downloaded locally, preventing developers from seeing how people are using them.

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.

AI scraping for LLM training data has significantly strained Wikipedia's infrastructure

From January 2024 to April 2025, the site's bandwidth increased by 50% as automated bots downloaded terabytes of data for the large language models powering AI tools. The Wikimedia Foundation found that bots accounted for 65% of the highest demand requests (e.g., videos) despite representing just 35% of page views.

Breaking down Anthropic's products and why their models are named after literary forms

Since March 2024, the Claude series of large language models includes three varieties—Haiku, Sonnet, and Opus—named to metaphorically illustrate increasing complexity and computing power. Claude Code can perform software engineering tasks for developers and, as of January 2026, writes nearly all of Anthropic's code alongside Opus 4.5.

Algorithms built to detect credit card fraud can deny charges if they perceive the attempted purchase is unfamiliar or unusual based on historical data, even when a person has enough credit to make the purchase.

Since fraudulent charges are rare compared with the much greater number of transactions across purchase histories, these models have been shown to be overly sensitive, lacking sufficient training data on what legitimate fraud looks like. Models also perform less reliably for groups in low-income areas or communities of color, given less available data from which to build an accurate baseline.

Help scientists at LIGO train AI to resolve glitches in observational data

Volunteers are provided with visual representations of signal abnormalities at various frequencies and tasked with categorizing them based on reference examples. Progressing through the tool unlocks more difficult images and additional features. Categorized images will be used to train machine learning algorithms for automated future classification.

The history of the Linux Foundation, a technology consortium supporting open-source projects

The nonprofit, founded in 2000 by IBM, HP, Intel, and others as the Open Source Development Labs, was established to build the infrastructure needed to scale with the growing development of Linux. Upon merging with the Free Standards Group in 2007, it rebranded as the Linux Foundation and, as of 2024, provides a "neutral home for code and collaboration" to almost 1,000 projects.

What are climate models?

Climate modeling what is a global climate model? What are their uses? Why do we believe them? What do they agree on? Why do they disagree? How do we improve them?

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.

Learn more about Palantir's core products, Gotham and Foundry

Palantir's core products, Gotham and Foundry, are designed to help government agencies and businesses turn complex or messy datasets into usable information. Gotham is designed primarily for government clients and connects disparate data points into recommendations and analysis. Foundry is designed for enterprise users.

How facts, hypotheses, theories, and laws in science differ from everyday language

Facts are observations of the world, and they can be pooled together for a given phenomenon to develop a hypothesis—a proposed explanation consistent with observations that requires further investigation. Experiments and additional data can rule out many hypotheses or graduate a proposed explanation into a widely accepted and validated theory.

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.

Explore more than 7,000 of the world's languages

Google collects open-source data on languages, detailing how many known speakers each language has, whether its endangered, and where in the world it is spoken.

South American lungfish has the largest animal genome, with 91 billion base pairs

This genome is 30 times larger than a human's due to significant amounts of "junk DNA" and jumping genes, or transposons. Transposons can jump or copy over to other parts of a genome and may aid adaptation.

Data center types vary based on location and the workloads they manage

Enterprise data centers are built on-premises and afford businesses greater control over the management and security of their data. Cloud data centers are built off-site and share their resources with millions of users through an internet connection, with the largest—hyperscale data centers—spanning millions of square feet. Managed data centers offer off-site hardware that companies can lease.

Quantum computers can simulate nature more effectively than classical computers

Ordinary computers struggle to simulate quantum fields due to entanglement, but systems built from those elements can model the complexity. Systems using particle in more than two states—qudits—have successfully modeled 2D electromagnetic fields and done so more efficiently.

How Amazon operates a 'flywheel' business model, a self-reinforcing cycle that drives continuous growth

Larger animals like giraffes, elephants, and blue whales have developed a unique defense against cancer, a phenomenon known as Peto's Paradox

As of 2025, the most cited papers in science include a review of graphene, methods to improve image recognition, streamlining electron density calculations, and optimizing protein studies.

Explore our favorite visualization of the immense scale of space, from an astronaut to the size of the observable universe

Massive 3D printers have been used to build homes by stacking layers of 'concrete toothpaste,' a fire-resistant material that effectively insulates against outside noise.

In 2025, four executives from large tech companies were given positions in the US Army Reserves—including Palantir Chief Technology Officer Shyam Sankar, which reinforced Palantir's ties with the US military and its role as a leading defense partner.

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.

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

In 2019, a laptop loaded with six of the worst malware in history and described as 'an exhibit of historical weaponry' sold for $1.3M.

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

The United States is one of only around two dozen countries with a fully federal system of government, in which power is divided between national and state governments.

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.

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

Radioactive elements can introduce software errors because the ionizing radiation they produce can detach electrons from atoms, and computer code relies on the distribution of electrons.

Electrical grids worldwide must double their power delivery capacity by 2050 to meet surging demand from AI and electric vehicles.

Venture capital is a form of funding provided to early-stage companies with high growth potential, typically in exchange for an equity (or ownership) stake

Power lines are dotted with large, brightly colored balls to improve the visibility of these thin wires to helicopters and small planes and help prevent potential crashes.

The use of fungi to detoxify land and water—a process known as mycoremediation—has been used to clean up natural resources affected by the 2025 California wildfires.

See how Budweiser's famous Clydesdales spend their days at Warm Springs Ranch, where roughly 70 to 80 of the beer brand's giant horses can be found at any given time.

The Nasdaq's 'Magnificent 7' companies are responsible for roughly 40% of the Nasdaq 100 index's gains, while the remaining 93 companies account for the other 60%.

See 33 images that became central to huge scandals, including a young Charles and Camilla, Michael Jackson's baby on a hotel balcony, and Hugh Grant's mugshot.

'Grand Canyon of the Yellowstone,' a work by American painter Thomas Moran, inspired Congress to establish Yellowstone National Park.

Ebola is a filovirus, meaning it resembles a thread under a microscope, though it sometimes twists and can resemble a bowl of spaghetti.

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