Meta challenges ChatGPT with code writing AI

  • Meta Platforms has announced a new AI-related product in a model that can write code for users.
  • Similar to ChatGPT, Code Llama can generate code and be used for debugging in Python, C++, C#, Javascript, and more with only prompts.
  • The company has been rushing to catch up with the rest of the tech industry in terms of AI products.

Social media powerhouse Meta is set to continue its push into generative AI after announcing on Thursday that it is working on a large language model (LLM) that can assist developers in writing computer code. The company recently shifted to developing AI-based features for its platforms such as Facebook, and is also working on ChatGPT-like chatbots for their users.

According to an official blog post, the code-writing model will be called Code Llama, and will be made freely available. It works the same as OpenAI’s chatbot in that you can generate code after writing prompts.

“It has the potential to make workflows faster and more efficient for developers and lower the barrier to entry for people who are learning to code. Code Llama has the potential to be used as a productivity and educational tool to help programmers write more robust, well-documented software,” Meta explains.

The firm says that Code Llama will be able to generate code using code and natural language as prompts, be used for code completion and debugging and supports Python, C++, Java, PHP, Javascript, C#, Bash and more.

“Programmers are already using LLMs to assist in a variety of tasks. The goal is to make developer workflows more efficient so that they can focus on the most human-centric aspects of their job, rather than repetitive tasks,” it adds.

Code Llama in action, according to Meta, writing code and plotting graphs after prompting.

But how does it match against OpenAI’s model? Well, Meta said it used two benchmarks to evaluate Code Llama’s performance against other popular LLMs that can write code. The company claims that Code Llama “performed better than open-source, code-specific LLMs” and outperformed its own Llama 2 model, which is not fine-tuned for coding.

Meta’s own data shows, however, that the leader among the models that can write code is GPT-4 according to the benchmarks. This model however is locked behind a subscription service from OpenAI, while Meta’s Llama Code is free to use.

The swing into AI by Mark Zuckerberg-owned company is a departure from efforts into the metaverse, a mixture of virtual and augmented reality technology. Since 2020 Meta’s metaverse division, Reality Labs, has lost over $30 billion. Meanwhile, the AI boom continues to pay dividends to companies that leapt towards chatbots that could simulate the writing of humans.

Recently, Nvidia announced record-setting profits on the back of selling processors that can run these LLMs for companies around the world with revenues of $13.5 billion.

Earlier in August Meta revealed that it was bringing ChatGPT-like chatbots to its social media platforms that could assume different “personas.” For example, if you select the Abraham Lincoln Persona, the chatbot will reply with the cadence of the long-dead president.

“Their purpose will be to provide a new search function and offer recommendations, as well as being a fun product for people to play with,” reported the Financial Times of the Personas.

Despite all the work that Meta is announcing in the sector it has yet to widely launch any substantial AI features for its platforms, outside of an image generator. You can petition Meta to try out Llama Code right now.

[Image – Muhammad Asyfaul on Unsplash]


About Author


Related News