GitHub utilizes AI, utilizes Microsoft Copilot to automate tasks and transform coding sector

Software developer Nikolai Avteniev was quick to recognize the potential of Microsoft Corp.’s Copilot coding assistant when he obtained a preview version in 2021.

Developed by Microsoft’s GitHub coding platform and based on OpenAI’s version of generative artificial intelligence, the assistant was not perfect and sometimes went wrong. But Avteniev, who works at ticket seller StubHub, was surprised by how cleverly it completed lines of code with just a few prompts. All he had to do was press the tab key and Copilot did the rest.

“Instead of using 15 keystrokes, it took three,” he recalled recently. “It was a nice little speed boost.”

Three years later, and now powered by the latest version of OpenAI’s GPT-4 technology, GitHub’s Copilot can do a lot more, including answering questions from engineers and converting code from one programming language to another. Because of this, the assistant is responsible for an increasingly large part of the software being written, and it is even used to program critical systems of companies.

Along the way, Copilot is gradually revolutionizing the working life of software engineers – the first professional group to use generative artificial intelligence en masse. Microsoft says Copilot has attracted 1.3 million customers to date, including 50,000 companies from small startups to the likes of Goldman Sachs, Ford and Ernst & Young. Engineers say Copilot saves them hundreds of hours a month dealing with tedious and repetitive tasks, giving them time to focus on more thorny challenges.

Bought by Microsoft in 2018 for $7.5 billion, GitHub dominates its market, betting that Copilot has the AI horsepower to battle rival services including Tabnine, Amazon’s CodeWhisperer and Google-backed Replit Ghostwriter. GitHub’s AI assistant is also a kind of beta test for the many other Copilots Microsoft is planning for Office, Windows, Bing and other lines of business.

As with AI in general, GitHub Copilot has limitations. The developers say it sometimes fetches outdated code, provides useless answers to questions, and creates suggestions that are buggy or may infringe copyright. Because the tool is trained on public and open code repositories, engineers are at risk of repeating security problems or introducing new ones into their work, especially if they blindly accept Copilot’s recommendations.

GitHub stresses that the tool is an assistant, not a replacement for human programmers, and has put the onus on customers to use it wisely. Strict guidelines are needed to prevent lazy programmers from simply accepting Copilot’s suggestions, said GitHub CEO Thomas Dohmke. He expressed his confidence that engineers would keep each other honest.

“The social dynamics of the team ensure that those who cheat by accepting code too quickly and who don’t actually go through the process defined by the team don’t get that code into production,” he said in an interview. .

Generative AI is the latest in a long line of innovations that have changed computer coding over the years. In the last century, program compilers sped up software development by quickly translating commands into ones and zeroes that computers could understand. More recently, Linux favored open source coding, allowing programmers to leverage each other’s work rather than writing everything from scratch.

Coding assistants like GitHub’s Copilot could be even more revolutionary, as generative AI has the potential to automate large amounts of what software engineers do.

At the moment it mostly makes them more efficient. StubHub’s Avteniev, who also teaches software engineering at the City College of New York, says Copilot’s predictive power helps programmers stay “in the flow” because they no longer have to stop to look things up. Avteniev has been coding for over 20 years, but sometimes he too forgets programming languages – forcing him to waste time Googling them. “Copilot prevents you from having to leave your current coding process,” he said. “Even if it produces nonsense, it’s still easier to just accept doing it and then fix it yourself.”

After more than 15 years of development, Aaron Hedges burned out before Copilot arrived. Hedges works for ReadMe, a startup that helps companies create technical descriptions of their application programming interfaces, or APIs. Like Avteniev, he takes advantage of Copilot’s autocomplete function. “Because I’m a fairly senior engineer, I can look at it and say, ‘Oh yeah, that looks right.'” He also likes that he can ask questions without leaving his programming window. “I don’t have to go away and open a browser, which can be really disturbing,” he said.

A Copilot subscription costs $10 a month, which Hedges is happy to pay for himself. After work, he builds websites for Dungeons & Dragons fans. With a toddler and another baby on the way, free time is precious. “The two hours I get to code in the evening are very important to me,” he said. “The more efficient I can be, the better.”

Few tasks are more tedious than debugging software—a process that can take up to 50% of an engineer’s time. Figma, which helps developers design user interfaces for apps or websites, says Copilot can create bug-testing programs in minutes rather than hours. “That’s the real value of AI,” said Abhishek Mathur, the company’s vice president of engineering. “It doesn’t replace our work, but it frees up our time to develop creative solutions.”

Some companies have started using Copilot to generate code for critical systems. Brewer Carlsberg uses it to write code for an existing tool that helps sales staff plan, prepare and document sales calls. According to CIO Sarah Haywood, given Copilot’s limitations, the brewer uses its own quality assurance process to verify that the code it generates works as intended. Eventually, he said, companies may also outsource this task. “Over time, people will build more trust in AI,” he said. “I don’t think we should behave to check everything AI does, otherwise we’re not really adding any value.”

Last year, the Canadian University of Waterloo published a test to evaluate the accuracy of the technology. The researchers collected a dataset consisting of code snippets with known flaws and fixes for those bugs. The researchers asked Copilot to generate these exact snippets to see if it would spit out the buggy versions. The assistant repeated the incorrect version 33% of the time, less than a human. In a quarter of the cases, the AI spit out the code along with the fix. Copilot was generally better at avoiding basic errors than more complex ones, said Mei Nagappan, a computer science professor at the school and one of the study’s authors.

“The analogy here is that we’re living in the era of driver assistance right now, not yet in the self-driving phase,” he said.

Software engineers can be slow to change their work habits. Many welcome Copilot, but are wary of becoming too dependent on it. A recent study funded by GitHub found that developers accepted the assistant’s suggestions only 27% of the time.

Engineers can also quickly blame Copilot if something goes wrong. When Etsy’s site went down for a short time last October and December, some of the company’s developers applied to Copilot for the outage. Etsy confirmed the incidents, but denied that Copilot was responsible. “While we certainly understand that engineers can discuss how Copilot could theoretically play a role in outages or issues, we have no evidence that the tool has actually led to any impact on customers,” the spokesperson said.

Copilot is expected to improve dramatically in the coming years. GitHub is already rolling out improvements, including an enterprise version that can answer questions based on a customer’s own programming code. This should help new engineers get up to speed and allow coders to work faster. In the coming months, GitHub will also let engineers use their employer’s own code base to help auto-finish the programs they’re working on. This makes the generated code more customized and useful.

GitHub cannot afford to sit still. At least a dozen startups aim to disrupt the market. Some take advantage of new models that have dramatically increased the amount of data that code contributors can quickly tap into, making it easier to create entire programs. “An AI programmer who can see all of your code will be able to make much better decisions and write much more consistent code than one who can only look at your code through a roll of paper towels, a small amount at a time. ” said Nat Friedman, investor and former CEO of GitHub.

Friedman backs a startup called Magic AI, which plans to create a “superhuman software engineer.” Peter Thiel-backed Cognition AI, on the other hand, is working on an assistant that can handle software projects on its own. This month, Princeton University released an open-source model for an artificial intelligence software design agent, and it seems like not a week goes by without another startup.

In interviews, few coders expressed fear of being replaced by AI. As in many industries, they say, automation frees them to focus on more challenging and interesting tasks. But Jensen Huang, CEO of hot artificial intelligence chipmaker Nvidia Corp., has a less rosy view. He recently predicted that coding as a career is doomed. Now that artificial intelligence makes it possible to code in plain English, Huang said anyone can become a programmer.