The latest update, which covers developments until August 7, 2023, highlights significant accomplishments and statements made in the fields of artificial intelligence, machine learning, and data science. This week's major news involved Jupyter, Meta AI, Stack Overflow, Stability AI and Tabnine.
Jupyter Notebook Releases Generative AI Extension for JupyterLab
Jupyter introduced Jupyter AI, a project that adds generative artificial intelligence to Jupyter notebooks. It enables users to generate full notebooks from simple natural language prompts, explain and write code, correct errors, and summarize content. The program links Jupyter to large language models (LLMs) from several vendors, including AI21, Anthropic, AWS, Cohere, and OpenAI, which is supported by LangChain.
Jupyter AI gives users the freedom to select their favorite LLM, embedding model, and vector database to meet their individual needs because it was created with ethical AI and data protection in mind. Data transparency is guaranteed by the open source nature of the software's core prompts, chains, and components. Additionally, it stores metadata regarding model-generated content, making it easier to follow AI-generated code throughout the workflow. Importantly, Jupyter AI protects the privacy of user data and only communicates with LLMs upon request; data is never accessed or transmitted without the user's express authorization.
Meta Open Sources Audiocraft, a Framework for Generating Sounds and Music
Meta has open sourced its text-to-music generative AI, AudioCraft, for researchers and practitioners to train their own models and help advance the state of the art. AudioCraft is a framework to generate high-quality, realistic audio and music from short text descriptions, or prompts. AudioCraft is largely based on three distinct AI models: MusicGen, AudioGen, and EnCodec. Using music samples that are owned and licensed by Meta, MusicGen creates music from text-based inputs. While using publicly accessible sound effects, AudioGen creates audio from text-based inputs. True-to-life audio outputs are produced by the EnCodec decoder, enabling higher-quality music generation with fewer artifacts.
AudioCraft simplifies the processes of compression, generation, and music/sound creation. This enables users to enhance the present code foundation for more efficient sound generation and compression methods, showcasing the adaptability of AudioCraft. In essence, you can build upon an established framework rather than starting anew. Your starting point will be determined by the dataset's existing limitations.
Stack Overflow Jumps into Generative AI with OverflowAI
Stack Overflow announced its strategic plan to incorporate generative AI into its public platform, Stack Overflow for Teams. This innovative approach extends to new product domains, including an IDE integration designed to seamlessly merge the extensive expertise encapsulated within 58 million community-generated questions and answers into the very space where developers channel their concentration and accomplish tasks. This comprehensive effort is collectively referred to as OverflowAI.
The OverflowAI initiative consists of a number of projects, including improved AI search on both public and enterprise platforms. OverflowAI is not a single product. Additionally, there is a Slack integration and an OverflowAI Visual Studio code plugin available for businesses. Additionally, OverflowAI will aid with enterprise knowledge intake for Stack Overflow for Teams. The ultimate objective is to make it simpler for businesses and developers to locate and utilize the information they require.
Stability AI Launches StableCode, an LLM for Code Generation
Stability AI announced StableCode, its very first LLM generative AI product for coding. This product is designed to assist programmers with their daily work while also providing a great learning tool for new developers ready to take their skills to the next level. Three tiers of StableCode are being offered: a base model for general use cases, an instruction model, and a long-context-window model with a maximum support of 16,000 tokens.
The StableCode model draws upon an initial compilation of programming language data sourced from the open-source BigCode project, further refined and fine-tuned by Stability AI. In its initial phase, StableCode offers support for programming endeavors encompassing languages such as Python, Go, Java, JavaScript, C, markdown, and C++.
Tabnine Announced Tabnine Chat
Tabnine has recently announced the beta of Tabnine Chat to interact with Tabnine's AI models using natural language. The chat application works inside the IDE, allows organizations to train on permissive code only, and can run on isolated environment deployment.
Tabnine Chat is an enterprise-grade coding chat application that lets developers interact naturally with Tabnine's AI models. Unlike ChatGPT, it focuses on aiding professional developers in sizable projects. It operates within the IDE, supporting both initial app creation and ongoing work, particularly in larger commercial projects. Tabnine Enterprise users can connect their repositories to benefit from internal project-specific assistance. The tool complies with security needs and offers isolated deployment options, ensuring code privacy. Tabnine Chat's training exclusively involves open-source code with permissive licenses, preventing contamination from copyleft code.