In a bold move that could potentially disrupt the AI field, Meta Platforms is gearing up to unveil a software designed to assist developers in automatically generating programming code. This innovative product, known as Code Llama, is set to challenge proprietary software from tech giants like OpenAI and Google, according to insiders privy to the development.
Code Llama
Code Llama, an open-source artificial intelligence model, is expected to launch as early as next week according to sources close to the development of the code writing AI. This new coding model is a direct competitor to OpenAI’s Codex model and is a significant enhancement of Meta’s Llama 2 software. Llama 2, also open-source, revolutionized the AI field by simplifying the process for companies to create their own AI apps without the need to purchase software from OpenAI, Google, or Microsoft.
The introduction of Code Llama is expected to further simplify the process for companies to develop AI assistants that can automatically suggest code to developers as they type. This could potentially divert customers from paid coding assistants like Microsoft’s GitHub Copilot, which is powered by Codex.
In a related development, Meta has also been working on CodeCompose, an AI-assisted code authoring tool that has been deployed internally. CodeCompose is based on the InCoder large language model (LLM) that combines generative capabilities with bi-directionality. This tool has been scaled up to serve tens of thousands of developers at Meta, across multiple programming languages and coding surfaces.
Other articles you may find of interest on Llama 2 recently released by Meta AI :
CodeCompose
The deployment of CodeCompose has presented unique challenges in terms of user experience and metrics in large-scale industrial settings. However, Meta has been able to address these challenges through careful design decisions about the model and system architecture.
The impact of CodeCompose on Meta’s internal code authoring experience has been significant. Over a 15-day period, CodeCompose made 4.5 million suggestions, with an acceptance rate of 22% across several languages. Furthermore, 8% of the code typed by users of CodeCompose was through accepting code suggestions from the tool. The tool has received an overwhelmingly positive reception, with 91.5% of users providing positive feedback.
In addition to assisting with code authoring, CodeCompose has also had other positive side effects such as encouraging developers to generate more in-code documentation and aiding them in the discovery of new APIs.
Meta’s foray into AI-assisted code generation with Code Llama and CodeCompose is a game-changer. It not only challenges established players like OpenAI and Google but also promises to revolutionize the way developers write code. As always assumed more information is made available we will keep you up to speed as soon anything is officially announced by Meta AI.
Source: Arxiv : TheInformation
Filed Under: Gadgets News
Latest togetherbe Deals
Disclosure: Some of our articles include affiliate links. If you buy something through one of these links, togetherbe may earn an affiliate commission. Learn about our Disclosure Policy.
