Microsoft launched the GitHub Copilot, a revolutionary text-to-code tool that relies on Artificial Intelligence (AI) earlier this year. The tool helps people code more efficiently, and the response has been phenomenal. Several social media users have been posting about GitHub Copilot and showering praises on the same for saving their time and effort. Let’s see why this Microsoft tool is so popular, and if could have some risks or dangers.
The world of AI is growing by leaps and bounds. Some algorithms can create detailed works of art from mere descriptive words. Specifically speaking, “Generative AI” is gaining popularity as it helps people create content that was once possible only by professionals who have spent years or decades honing their skills.
Earlier this year, Microsoft introduced Generative AI to the world of coding. Called GitHub Copilot, it is a tool that is supposed to work alongside computer programmers. Technical jargon aside, GitHub Copilot suggests blocks of code that might follow what the programmer is typing.
Simply put, Microsoft’s GitHub Copilot is an autocomplete program, but instead of suggesting words in a sentence, the program suggests computer code. Microsoft has trained the AI program to “speak” in Python and JavaScript, which are two of the most popular coding languages used today.
GitHub Copilot is exceptionally useful for repetitive or common tasks. In other words, programmers can use the program for filling in commonly or frequently required blocks or chunks of code.
In the initial stages, GitHub Copilot can confidently suggest code that is necessary but not complicated. The program isn’t creative, but with millions of lines of code, it could replicate and predict patterns in the future. Given enough training, GitHub Copilot could also get a little imaginative and suggest better and more efficient code.
Microsoft claims GitHub Copilot has become quite popular and is getting better. The company mentioned an average of 26 percent of all suggestions made by Copilot were accepted. Surprisingly, Microsoft has trained the AI program in natural language too. This means GitHub Copilot could work for programming and human languages.
GitHub Copilot is surely a promising platform. Any AI program that reduces human effort will obviously be welcomed with open arms. The problems could start if developers rely on the program extensively or blindly.
Although it is a little early to say, GitHub Copilot may have some inherent biases which are common among humans. Any AI program needs a lot of data, and “large language models” such as this, would require enormous amounts of data, and that too in a short time. Microsoft may have relied on data that might not be well documented and vetted to remove biases.
Incidentally, Microsoft has created and deployed a content filter for GitHub Copilot and other partner programs. The company claims it can filter harmful content, which might include hate speech or racism. Additionally, Microsoft works with human moderators who monitor the program, confirmed Sarah Bird, who leads the product development of Responsible AI tools for Microsoft’s Azure AI.
Microsoft’s tool is certainly helpful, but there have been some complaints about it causing harm to the open-source community. Additionally, Microsoft may have to deal with some legal trouble as the tool could easily draw from an open-source codebase to train its ML models without the authors’ permission. In fact, the Free Software Foundation (FSF) has already branded the project “unacceptable and unjust”.
Microsoft seems to have intentionally positioned GitHub Copilot as an assistive technology and has insisted that it will go ahead with the program. The company stresses the platform can help with repetitive work, leaving coders to work on code that needs imagination and creativity.
Leave a Reply