Meta Debuts AI Code Creation Tool, ‘Code Llama’

Meta, formerly known as Facebook, announced the launch of its new, free state-of-the-art AI tool that is able to take text prompts that streamline the process of code creation and debugging, according to a company press release on August 24th. 

Named ‘Code Llama,’ the tool leverages Meta’s Llama 2 large language model (LLM) to essentially make workflows faster and more efficient for developers and for those beginners who want to learn how to code. 

Designed as a free educational tool to help programmers write “more robust, well-documented software,” Code Llama is able to generate code and natural language about the code from both the code itself and natural language prompts such as “write me a function that outputs the “fibonacci sequence.” It can also reportedly be used for commercial purposes without a fee. 

One of its more notable features is its adaptability to different scenarios – it can either construct code segments from these text prompts or identify and rectify issues within specific code strings (i.e. debugging). 

According to Meta, Code Llama emerges from the growing trend of programmers utilizing LLMs to enhance various aspects of software development, ranging from crafting new applications to troubleshooting existing codebases. 

Meta’s strategy involves releasing Code Llama in three different sizes, catering to various project requirements. Notably, even the smallest variant is designed to operate on a single Graphics Processing Unit (GPU), making it ideal for projects that necessitate low-latency responsiveness.

Other Code Generators

In addition to Meta’s code generator tool, we have started to see other types of code generators sprout – however, the issue surrounding copyright infringement continues to be a major setback. 

Last November, a class-action lawsuit against Github’s parent company Microsoft, as well as OpenAI, for allegations of having violated copyright law with GitHub’s ‘Copilot’ tool that is powered by OpenAI’s GPT-4.

Google’s DeepMind team has also teased its coding tool, AlphaCode, but has not yet released it to the general public. It also is currently testing an internal AI tool that supposedly will be able to provide individuals with life advice and the ability to perform at least 21 different tasks.

Similarly, Amazon Web Services (AWS) uses ‘CodeWhisperer’ to write and debug code. 

Turning to Meta, the company has gone a step further by releasing specialized iterations of Code Llama. One such version is Code Llama-Python, tailored to Python programming, and another is Code Llama-Instrct, designed to comprehend natural language instructions. It’s important to note that these distinct iterations are not interchangeable. The company explicitly advises against employing the basic Code Llama or Code Llama-Python for interpreting natural language instructions.

Having said that, while Meta’s newest innovation does position it as a formidable contender, conversations surrounding ethical use will continue to surface as the AI landscape continues to evolve.