GitHub’s incorporation of artificial intelligence tools into its platform, designed to help developers improve their productivity, is key to the platform’s strategy to grow its user base tenfold – from 100 million to a billion developers in the next decade.
According to GitHub CEO Thomas Dohmke, who spoke to TechCentral at the Africa Tech Festival in Cape Town last week, AI development tools are not only boosting productivity for professional developers but also making code more accessible to students, hobbyists and even children who want to learn to code.
“We have set ourselves the aspiration of growing from just over 100 million to helping a billion humans become software developers,” said Dohmke.
“That does not mean they all become professional software developers in the same way that not every child that learns music, art and physics in school becomes a musician, an artist or a physicist. Coding is a form of literacy, and I think everybody should have the fundamental skills to create software.”
As a Microsoft subsidiary, GitHub played a big role in the development of Microsoft’s AI tool, dubbed Copilot, which is based on OpenAI’s large language models. Microsoft has invested US$13-billion in OpenAI since 2019.
Last month, GitHub announced the addition of “multimodal” functionality to its AI-infused code completion and programming tools, which will allow developers to choose models other than OpenAI’s in their workflows. Newly added models include Anthropic’s Claude 3.5 and Google’s Gemini 1.5 Pro.
Choice
According to Dohmke, the move to expand GitHub’s AI-coding functionality beyond parent Microsoft’s tooling aligns with GitHub’s ethos of giving as much choice to developers as possible. He said programmers will have to play with the various models in different scenarios, as they would with any other tool or framework, to discover which models are best suited to their work.
Code completion was the first – and remains the most common – use case for AI in programming. But as Dohmke puts it, there are many stages in the software development life cycle, and AI has a role to play in all of them.
Read: Microsoft adds AI chatbot to GitHub dev tools
GitHub has a chat function that allows developers to use AI to help them understand the code they are working with. The function is useful for students exploring an open-source code base, new developers getting to know the code their organisation uses, and even experienced developers who switch between multiple projects within an organisations and must understand a new code base quickly.
“Moving from one team to another, or from one company to another, happens often in the software space. Sometimes within the same project there are dependencies on other code that also need to be understood. With AI, you can basically look at the code base, have it explained to you, navigate it and understand how it is designed,” said Dohmke.
While the benefits of having AI assist in explaining complex code design are largely used by enterprises today, Dohkme sees this functionality having a profound impact on how programming is taught, especially when it comes to opening the craft up to those who in the past have found it inaccessible.
For the African population, the ability to use voice commands to interact with GitHub’s chat function and have it explain code in multiple languages, including Afrikaans, for example, will expose programming to much wider audiences, he said.
Regardless of the tasks being performed, however, learning how to use AI tools to boost productivity and maximise output is a skill that is fast becoming essential to knowledge workers all over the world. Dohmke said the incorporation of AI into the teaching environment, instead of viewing its use by students as cheating, will equip learners with the skills to prompt correctly and even identify when the AI is “hallucinating” (producing erroneous output).
Read: How SA educators are fighting the AI cheating scourge
“As a CEO of a company and an engineering leader, I want my engineers to leverage AI to its full potential. But that in turn means I want new employees, whether they are coming from university or from other companies, to understand Copilot before they even apply for the job. How to prompt a model to get the best outcome is going to become a key skill for knowledge work,” said Dohmke. – © 2024 NewsCentral Media
Get breaking news from TechCentral on WhatsApp. Sign up here
