Connect LLMs and other AI systems to powerful, accurate and current computation and knowledge—through both natural language and high-level computational language.
Building on 35+ years of delivering tools for human R&D, education, enterprise and consumer applications, Wolfram technology has long been a component of intelligent assistants and other AI systems.
Wolfram makes ChatGPT smarter by giving it access to powerful computation, accurate math, curated knowledge, real-time data and visualization through Wolfram|Alpha and Wolfram Language. By expanding on our original plugin for ChatGPT, Wolfram GPT delivers broad and deep coverage from chemistry to geography, astronomy to popular culture, nutrition to engineering, along with algorithmic code execution.
APIs FOR AIs
Wolfram|Alpha API for LLMs
Wolfram|Alpha APIs have been used by leading intelligent assistants, websites, knowledge-based apps and other partners for more than a decade, answering billions of questions in hundreds of knowledge domains. The Wolfram|Alpha API for LLMs is now available, providing fast, reliable assistance with computational and fact-based tasks that large language models (LLMs) consistently struggle with. Designed to work off the shelf with leading LLMs and plugin and tool frameworks (ChatGPT, LangChain, etc.), the API for LLMs is designed to be directly called by LLMs—rather than end users—using simple, natural language queries and to return precise, trustworthy Wolfram|Alpha results in a format tailor-made for ingestion by LLMs.
Wolfram Language API for LLMs
For projects requiring more advanced computation and code execution, the Wolfram Language API for LLMs gives AI projects a direct link to Wolfram Cloud technology for evaluating Wolfram Language code and integrating sophisticated analyses and visualizations with LLM responses. The Wolfram Language API for LLMs has been optimized to recognize common coding errors and to autocorrect or suggest code improvements in its responses. With native support for functions related to mathematical computation, machine learning, graph and network analysis, image processing and much more, the Wolfram Language API allows LLM-based projects to immediately execute code for virtually any task in a single, unified environment that does not have dependencies on external libraries. API users also gain direct, programmatic access to the extensive computational data contained in the Wolfram Knowledgebase.
Wolfram API Capabilities for AIs
Natural Language Input
The Wolfram Natural Language Understanding (NLU) System, used by leading intelligent assistants for over a decade, allows AI systems to retrieve precise, disambiguated data and perform complex calculations through simple natural language queries. This is the primary method of interacting with the Wolfram|Alpha API, understanding everything from simple English-language questions to imprecise mathematical expressions. Wolfram NLU also enables the Wolfram Language API for LLMs to interpret entity names, quantities, date and time specifications and many more plain-English inputs and automatically transform them into computable Wolfram Language objects.
For large language model (LLM) "fact injection" and other AI-related use cases, Wolfram APIs provide fast and accurate results in various forms. Wolfram|Alpha APIs can provide results as plain-English sentences, step-by-step computational explanations, concise facts, visualizations and more. Wolfram Language has built-in capabilities for creating and deploying custom APIs on the web and elsewhere, allowing even more flexibility in integrating Wolfram capabilities with custom AI projects.
Both Wolfram|Alpha and Wolfram Language APIs enable AI systems to generate a wide variety of high-quality plots, graphics, maps, infographics and interactive visualizations. Wolfram|Alpha suggests useful visualizations to provide context for different inputs, while the Wolfram Language API grants access to all the visualization functions built into the language.
Both Wolfram|Alpha and Wolfram Language APIs offer LLMs access to thousands of reliable algorithms, covering numerical and symbolic mathematics, statistics, physics, computer science, cryptography, blockchains and more. Whether using natural language queries with the Wolfram|Alpha API or feeding code directly to the Wolfram Language API, LLMs can leverage these powerful computational capabilities.
The Wolfram Knowledgebase, with trillions of data points across hundreds of domains, serves as a continuously updated resource for LLMs and other AI projects. Backed by unambiguous semantic representations of entities and property values—such as quantities, dates, time series, geo polygons and images—the Wolfram Knowledgebase integrates seamlessly with Wolfram Language and serves as the primary repository of facts for Wolfram|Alpha. This allows AI systems to easily access authoritative data and perform accurate computations about the real world.
Computational Language Code
Train your AIs to write efficient, powerful and concise Wolfram Language programs by utilizing hundreds of thousands of human-annotated code examples.
Educate your AI about backgrounds, techniques, computational representations and visualizations of thousands of applications, mechanisms, effects, algorithms, theorems, natural laws, ... from the natural sciences, mathematics, engineering, technology, business, art, finance, social sciences and more.
Turbocharge the mathematical skills of your AI with millions of natural language questions translated into precise computational Wolfram Language representations to have it ace AP exams and mathematical olympiads.
For Use in Large-Scale Systems, Ask Us About...
Structured Computable Data Snapshot
For data that does not require related computation, snapshots of the Wolfram Knowledgebase may be made available for offline use. One-time or reoccurring deliveries can be arranged, from individual domains to near-complete snapshots of the non-computational parts of the Knowledgebase.
Mathematical Testing Data
Show off your LLM's math-PhD readiness on our forthcoming yearly AI symbolic mathematics competition. Test the calculus capabilities of your network on a unique test set of indefinite integrals especially created for LLMs by world-renowned integrator Oleg Marichev.
Tools for Humans Making AIs
Humans require access to cutting-edge frameworks to build advanced AI services. Wolfram technologies serve as a solid foundation for the development of sophisticated and reliable AI systems.
Wolfram Language is the world's only full-scale computational language—providing a computational representation for both real-world and abstract objects and processes, with broad built-in curated data, models, methods, algorithms and computational knowledge.
Machine Learning Framework
Wolfram Language offers fully automated and highly customizable machine learning functions to perform classification, regression, clustering and many other operations. Classical methods are complemented by powerful, symbolic deep-learning frameworks and specialized pipelines for diverse data types such as image, video, text and audio.
Neural Net Repository
The Wolfram Neural Net Repository is a publicly accessible resource that contains a growing assortment of neural network models, both pre-trained and untrained. These models can be readily utilized for a variety of purposes, including but not limited to evaluation, training, visualization and transfer learning.