Meta just released a coding version of llama 2
Code llama: A Code Generator for Efficient Embedded and Distributed Software Development (Code LLMs, Google, AlphaCode, and Google)
Meta claims Code Llama performed better than publicly available LLMs based on benchmark testing but did not specifically name which models it tested against. Code llama scored over 50 percent on the code benchmark HumanEval and was able to write code based on a text description.
“Programmers are already using LLMs to assist in a variety of tasks, ranging from writing new software to debugging existing code,” Meta said in a blog post. To make the developer process more efficient, they want to focus on the human aspects of their jobs.
Code generators have been helping developers work for a while now. The GPT-4 is the main component of Copilot, which was launched in March. GitHub Copilot can also rewrite old code to update it. Amazon’s AWS also has CodeWhisperer, which also writes, checks, and updates code. And yes, Google also has a code-writing tool in AlphaCode, but that isn’t out yet.
Copilot, a tool used to reproduce licensed code, is being sued by Microsoft and OpenEase for allegedly violating copyright law.
The release of Meta Llama 2 and the formation of Copilot: an AI-assisting platform for automated coding
“It’s exciting that they’re releasing the weights to the community,” says Deepak Kumar, a postdoctoral researcher at Stanford who has studied AI coding, referring to the parameters of the neural network at the core of the model.
Kumar says the release of Meta’s regular language model Llama 2 led to the formation of communities that discuss how it behaves and how it can be changed. It gives us flexibility to play with what is going on under the hood, compared to the closed-source models we are currently using.
Kumar says developers are likely to build new kinds of applications using Code Llama. Kumar’s own research has explored how Artificial Intelligence can sometimes cause less secure code, and if it is possible to create a programming assistant that performs various additional safety checks before advising a chunk of code. Kumar indicates that the release could lead to the creation of specific types of assistants for coding. He says that you can build a variety of tooling on the model.
In May 2021, GitHub, a subsidiary of Microsoft, launched Copilot, a plug-in for coding programs that auto-completes sections of code based on the first line or a comment typed by the user. A large language model called the GPT, used by Copilot, is a version of Open Artificial. That model is trained further, using code that can be found in GitHub, as well as by contractors who are paid to modify their own code.
Masad says that Meta is likely to limit the amount of training data that they use because of the lawsuit against them. Copilot costs $10 per month for individuals and $19 per month, per user, for businesses.
