Copilot: Emerging Insights on Generative AI and the Future of Software

Over the last few weeks, our team of engineers and technology architects have been evaluating Copilot - the Generative AI-powered code completion tool developed by GitHub and OpenAI. With Copilot now out of public beta and offering a paid subscription model, we decided it was time to test it out.

Although our team is kicking the tires and battle-testing the full potential of what Copilot can do, we’re bringing our early observations to the wider community. Here are four key insights worth sharing:

#1. Copilot won’t replace developers. If you're like us, you're embedded in the deep discussions and discourse that Generative AI and Copilot could eliminate the need for human coders. Our experience with Copilot shows that humans are firmly in the driver’s seat. Copilot is a competitive advantage that can help teams increase their efficiency, but it’s just a tool and cannot perform the job of today’s software engineer. Like similar advances in technology, we may find that Generative AI results in more jobs - not less.

#2. Automating repetitive tasks FTW. Copilot’s sweet spot is creating routine, boilerplate code – the kind of code most engineers prefer to delegate to a machine anyhow. This delegation of mundane tasks keeps developers “in the flow” and boosts productivity by enabling them to focus on the high-value, mission-critical aspects of their work.

#3. With great power comes great responsibility. Tools like Copilot are incredibly powerful, but we need to be deliberate about how and when we use them. As with any online tool or service, it is critical to ensure its use does not conflict with a business’ intellectual property requirements. While we are bullish on Generative AI’s potential to accelerate technological progress, we must be responsible stewards and avoid using Copilot in situations where it could inadvertently reveal sensitive intellectual property.

#4. A new era of software democratization is upon us. Since the early mainframes of the 1960s, software development has been the domain of the few who amassed the technical knowledge and programming proficiency required to bring an idea to life. With Copilot, the entire corpus of general programming knowledge can be made available through a simple prompt, removing many of the barriers that previously impeded newcomers. We’re encouraged this new influx of programmers, makers, and creators will be a good thing for progress and innovation.

OUR BOTTOM LINE: we're fans of Copilot. You tell it what you want, and it attempts to write code that does the thing you asked for. It's simple and it works. It's not perfect - and that's good enough for a lot of scenarios.

Time will tell if tools like Copilot will live up to all the hype. From our perspective, the future is looking bright.