In the October of 2021, developers who utilize the GitHub software development platform gained access to a remarkable AI tool called Copilot. Developed in collaboration with OpenAI, the company responsible for ChatGPT, Copilot employs advanced generative AI technology to generate computer code upon request, rather than text. The performance of Copilot is quite impressive, as GitHub estimates that it successfully provides auto-complete suggestions for coding about 50% of the time. This is expected to greatly enhance the productivity of the millions of developers who currently utilize the tool. Consequently, we are left pondering a fascinating question: just how impactful will Copilot prove to be?
Fortunately, the work of Thomas Dohmke at GitHub, Marco Iansiti at the Harvard Business School, and Greg Richards at Keystone.AI provides us with an answer. They have assessed the usage of Copilot among developers and determined that AI-powered coding is generating a substantial shift within the software industry, which is set to accelerate global GDP growth by over $1.5 trillion by the year 2030. They assert that this symbiotic relationship has the potential to shape the future construction of the world’s software, benefitting generations to come.
GitHub is a prominent software development platform that fosters seamless collaboration among developers on a global scale. Established in 2008, it was acquired by Microsoft in 2018 and currently boasts 100 million registered developers. GitHub has rapidly become the favored repository for software development, particularly for open-source projects. Copilot utilizes the same Generative Pre-Trained Transformer technology found in ChatGPT, but instead of generating text, it generates code. As a result, it holds significant power to influence software production. To quantify its impact, Dohmke and his colleagues examined the way nearly a million developers on GitHub utilize Copilot and the code it produces. They discovered that, on average, users accept almost 30% of code suggestions and report increased productivity as a result. It is particularly interesting to note that the influence on productivity grows over time, with less experienced developers reaping the greatest benefits. This aligns with previous research suggesting that AI assistants are most beneficial to those with less experience, as they have the most to learn. Furthermore, Dohmke and his team reveal that most of the innovation related to AI-powered coding occurs within the open-source project realm, predominantly led by individual developers rather than corporations. Their findings suggest that generative AI software innovation is driven by the open-source ecosystem, particularly in the United States.
So, what is the worth of this increased productivity? Dohmke and his colleagues point out that software development already makes a significant contribution to global GDP, but its potential is stifled by the global shortage of developers. They argue that the increased productivity resulting from Copilot has the potential to help close this gap, thereby adding $1.5 trillion to global GDP by 2030. However, they stress that this estimation is conservative, emphasizing the accelerating adoption of AI-powered coding techniques and suggesting that we are merely witnessing the beginning of an immense transformation. In conclusion, they assert that as more developers embrace these tools and become proficient in prompting generative AI, it is evident that this new approach to software development has established an irreversible connection between human beings and artificial intelligence. This connection will likely shape the future of software development for years to come. So, fasten your seat belts and get ready for a thrilling ride!
Ref: Sea Change in Software Development: Economic and Productivity Analysis of the AI-Powered Developer Lifecycle: arxiv.org/abs/2306.15033