GitHub Opens Up Its AI-Assisted Coding Tool to the Public
By Adam Pease
GitHub just announced the release of its Copilot artificial intelligence solution for developers. The tool, which has been in a closed beta for some time, will now be publicly available for $10 a month. This blog discusses the implications of the release for GitHub’s parent company, Microsoft, and surveys the rise of the generative content market.
What Is GitHub Copilot?
Github Copilot is an AI-assisted tool for programmers that suggests and auto-completes code based on the development context. Copilot is the product of GitHub’s collaboration with leading algorithm provider OpenAI. With OpenAI being a Microsoft partner and GitHub being a Microsoft company, this move represents a new push by Microsoft to establish itself as a leading provider of AI services to the enterprise. It dovetails with Microsoft’s recent announcement of its Azure OpenAI Service, which offers enterprises access to the same large language models that were used to develop Copilot.
Copilot represents one example of what Aragon has called generative content, an emerging dimension of the workflow and content automation market. Generative content refers to content that has been produced with AI algorithms—in its current incarnations, it ranges from image, video, to text, and, in the case of Copilot, code. Copilot, based on OpenAI’s Codex language model, itself trained on a large corpus of GitHub code, has impressive abilities when it comes to supplementing the work of programmers. And yet, it is still not capable of supplanting a human employee—one analysis found that nearly half of Copilot’s code had security vulnerabilities.
Generative Content and the Future of Work
Copilot is an early example of the way generative content will transform the future of work. Instead of outright replacing human employees, Aragon predicts that AI will emerge as a supplement to human knowledge workers, helping automate simpler aspects of workflow to free up human labor for more complex and intellectually rigorous work. In the case of coding, generative content has immense potential to accelerate the development life cycle, especially when it comes to repetitive programming tasks like software testing.
While Copilot has demonstrated impressive generative abilities, its results are inconsistent. Developers have complained that it performs poorly at some tasks even as it excels at others. It remains to be seen whether using AI-assisted tools like Copilot will emerge as a best practice amongst developers. However, many coders are enthusiastically expressing how much efficiency Copilot has brought to their work, and how the tool seems capable of anticipating their ideas before they type them. There is no doubt that Microsoft is making strides by making its tool available to the public.
Will developers adopt GitHub Copilot for its potential to accelerate their workflow, despite its security vulnerabilities and inconsistencies? It’s an open question. Microsoft is working out Copilot’s kinks though and has now made itself the player to beat in the assisted coding market. Copilot’s release does suggest we are on the edge of a revolution in generative content.