Wallstreetcn
2023.08.18 18:26
portai
I'm PortAI, I can summarize articles.

Open Source Beats Paid Ecosystem? Meta Platforms to Launch "Free Version of Programming Tools," Targeting OpenAI

Meta Platforms will launch an open-source AI software as early as next week, aiming to assist developers in automatically generating programming code. This bold move, following the success of the Llama 2 language model, has the potential to disrupt the field of artificial intelligence.

Meta Platforms is expected to announce another "blow" to OpenAI as early as next week, according to two sources familiar with the matter.

The company will launch software that helps developers automatically generate programming code. This open-source programming AI model, called Code Llama, directly challenges the paid proprietary software of OpenAI, Google, and other companies.

Code Llama will "benchmark" OpenAI's Codex model and significantly enhance Meta Platforms' open-source language model, Llama 2. The open-source nature of Code Llama will make it easier for enterprise users to develop their own AI assistants, which can automatically suggest code as developers type and compete with paid programming assistant tools supported by Codex, such as Microsoft's GitHub Copilot.

Analysts believe that enterprise users may prefer to use open-source programming models to develop their own programming assistants in order to protect their source code. Generating automatic code suggestions has always been one of the most popular use cases for large language models. A year ago, GitHub started charging developers of Copilot $10 per month, and recently, many emerging programming assistant companies have received funding from venture capital firms such as Sequoia Capital.

The current consensus in the comment community is that the launch of another open-source AI model represents Meta Platforms' latest bold move to disrupt the field of artificial intelligence. The open-source release of Llama 2 has shocked the industry, as it simplifies the process for companies to create their own AI applications without having to purchase related software from OpenAI, Google, or Microsoft, thereby completely changing the AI landscape. It seems that the "AI battle" between Meta Platforms and the highly popular OpenAI is intensifying.

According to a report by Wall Street News, Meta Platforms may launch a series of AI-driven chatbots with different characteristics as early as September, directly competing with OpenAI's ChatGPT. OpenAI is also working on the development of its open-source language model, G3PO, but the release date is still unclear. Surprisingly, Microsoft, which heavily funded OpenAI, partnered with Meta Platforms to release the commercially available version of their open-source AI model, Llama 2, offering an alternative to both OpenAI and Google models.

On one hand, open-source models like Code Llama enable newcomers in the industry to participate in the competition more quickly, and it also allows large companies concerned about source code security to build their own internal models, making external paid providers redundant and directly disrupting the industry landscape.

At the same time, Meta Platforms' employees will also use the same open-source models to develop consumer products. As more and more developers worldwide use Meta Platforms' open-source models, they can discover security flaws and errors, or make improvements in other ways, which also benefits Meta Platforms. According to some analysis articles, Meta Platforms revealed its internal AI programming assistant CodeCompose in May this year. Meta Platforms stated that this programming model is trained based on programming data, which means it will avoid potential copyright issues encountered by other AI programming assistants. However, there are no plans to release a public version of CodeCompose yet.

CodeCompose is based on the InCoder large language model, combining generation capabilities with bidirectionality. It has provided automatic code generation services across multiple programming languages and interfaces to tens of thousands of developers within Meta Platforms.

It is reported that CodeCompose has had a significant impact on the internal code writing experience at Meta Platforms. In just 15 days, CodeCompose proposed 4.5 million representative suggestions, with a user acceptance rate of 22% across multiple programming languages, and nearly 92% of users provided positive feedback. In addition to assisting with code creation, CodeCompose also encourages developers to generate more code documentation and helps discover new APIs.

Therefore, some people assert that Meta Platforms is entering the field of AI-assisted code generation with Code Llama and CodeCompose, "changing the game rules." It not only challenges industry giants like OpenAI and Google but also has the potential to completely change the way developers write code. On Friday, August 18th, Meta Platforms' stock price decline narrowed during the trading session in the U.S. stock market.