At Awtana, we pride ourselves on always being at the forefront of the latest trends and news in technology. Today, we want to share exciting news that is making waves in the world of artificial intelligence and software development. Mistral, the French AI startup backed by Microsoft and valued at $6 billion, has launched its first generative AI model for coding, named Codestral.
In this blog post, we will explore what makes Codestral special and how it could impact the future of software development.
Codestral is an AI model designed to help developers write and manage code. It was trained on more than 80 programming languages, including Python, Java, C++, and JavaScript, as Mistral explains in a blog post. This model can not only complete coding functions and write tests, but it can also "fill in" partial code and answer questions about a codebase in English.
However, Codestral's openness is a subject of debate. Although Mistral describes the model as "open", the license prohibits the use of Codestral and its outputs for commercial activities. There is an exception for "development," but even that has its limitations: the license explicitly prohibits "any internal use by employees in the context of the company's commercial activities." The reason behind these restrictions might be that Codestral was trained in part on copyrighted content. Mistral neither confirmed nor denied this in its blog post, but there are indications that the startup's previous training datasets contained copyrighted data.
In any case, Codestral may not be worth the effort. With 22 billion parameters, the model requires a powerful PC to run. (Parameters essentially define an AI model's ability to analyze and generate text.) And while it outperforms the competition according to some benchmarks (which, as we know, are not always reliable), it is not a landslide victory.
Despite these limitations, Codestral will certainly fuel the debate over the wisdom of relying on generative code models as programming assistants. Developers are adopting generative AI tools for at least some coding tasks. In a June 2023 Stack Overflow survey, 44% of developers said they use AI tools in their development process, while 26% plan to do so soon. However, these tools have glaring flaws.
An analysis of more than 150 million lines of code committed to project repositories over the past few years by GitClear found that generative AI development tools are resulting in more code errors in codebases. In addition, security researchers have warned that these tools can amplify existing errors and security issues in software projects; according to a Purdue study, more than half of the answers provided by OpenAI's ChatGPT to programming questions are incorrect.
Despite these problems, companies like Mistral and others will continue trying to monetize (and gain market share with) their models. This morning, Mistral launched a hosted version of Codestral on its conversational AI platform Le Chat, as well as its paid API. Mistral has also worked to integrate Codestral into application frameworks and development environments such as LlamaIndex, LangChain, Continue.dev, and Tabnine.
At Awtana, we are always excited about technological innovations and their potential impacts on our lives and work. Mistral's launch of Codestral represents a significant breakthrough in the field of generative AI applied to software development. While the model has its limitations and challenges, its ability to assist in coding and codebase management is impressive. However, it is also crucial to address and consider the ethical and practical implications of its use. We will continue to closely monitor these developments and share the latest news and analysis with you.
