AI will not replace developers, will instead create more work: GitHub CEO Thomas Dohmke

GitHub, which was acquired by Microsoft for $7.5 billion in 2018, is now working on infusing AI across every step of the developer lifecycle. This includes a code completion tool Copilot in 2022, Copilot Chat for natural language-powered coding in 2023 and an AI developer environment GitHub Copilot Workspace in April 2024.

With the latest advances in AI and tools like GitHub Copilot, Dohmke believes that natural language will be the universal programming language in the future, which will allow anyone to code in the language they speak.

He also says that AI will not replace developers but will instead create more work.

Edited Excerpts

You've been in India since last week. What are you seeing on the ground as far as AI is concerned?

I've seen lots of energy in both Delhi and Bengaluru, I met with a number of customers such as Infosys and Paytm that are using GitHub. There was also tremendous excitement from developers to come to our event.

I think that's part of the reason for this trip, which is that, on the one hand, India is growing fast in the developer community. We are actually predicting that by 2027, India will be the largest developer community on Earth, when looking at the number of GitHub accounts.

Today, the United States is first and India second, but if you look at the growth rates in India being already at over 15 million developers. We can forecast that it will overtake the US by 2027.

Looking at the computer science graduates in India, you're probably not going to lose that number one spot for a long time after that. There's a great rise of developer power here.

On the other side, we have Generative AI, that's in everybody's mind, This year, we see it as the year of AI adoption and so you have those two momentums coming together. I call it the great convergence.

You have a new technology that, at least for software developers, is the biggest step function since we started coding in 1950s and 1960s when we moved from punch cards to compilers. At the same time you have an incredible workforce that's highly educated.

Can you give us a sense of how big the GitHub community is in India today?

In the overall GitHub community, we have over 100 million accounts and as of last month (May 2024), India is at 15.4 million developers. Among the top communities, India is the fastest growing, growing faster than the United States for various reasons, like the sheer population of India and the high number of students in computer science and engineering skills.

These 15.4 million developers, are these students, or product managers working in startups or are they enterprise accounts?

We have a large population of students, like in the United States and in Europe, who get access to GitHub Copilot for free. These are kids at age 9, 10, 11 who are starting to learn to code because they have Python classes in school or they're just interested.

So students are a large community here in India. Startups are also a large community. And then it's the commercial sector comprising companies across a wide variety of industries including the likes of Infosys, Paytm, Makemytrip, and Cognizant.

Read: Indian developers are second biggest in generative AI project contributions on GitHub

Can you talk to us about GitHub’s India presence?

We're part of Microsoft which has over 20,000 employees in India. Many of them are engineers in Hyderabad, and also here in Bengaluru. We are closely collaborating with the Microsoft Research team in Bengaluru on the future of Copilot and Copilot workspace.

Talking of Microsoft, a few months back when Satya Nadella was visiting India, he said that he wanted Microsoft to 'copilot' India's AI journey. What role do you see GitHub playing and what has GitHub's own adoption journey being like in India.

At GitHub, we are the earliest adopter of Copilot ourselves. We started using Copilot about six months after we had access to OpenAI's GPT-3 in June 2020.

So we now have three years of experience with using a Copilot to build software. At the end of the day, we are very similar to many of the software companies here in India. We are building software. And we are living through the same pain that they are living through, the endless number of backlog items, huge demands, we are asking all our developers to move fast and do more, while we also have increased the amount of work through all these regulations, and compliance etc.

We are now bringing this into India in partnership with Infosys, where we opened the center of excellence, where the idea really is that both GitHub and Infosys employees working together on site on their campus to teach and skill Infosys employees to use Copilot to understand prompting and enable them in order to bring those skills into their day-to-day story and their customer relationships.

Last year, GitHub had to let go of its entire India engineering team. How are you now approaching the company’s operations in India?

The layoffs were a very hard decision that we had to take, given the economic environment last year and also our priorities. But the majority of these engineers got re-hired into Microsoft right after.

Moving forward, we rely on Microsoft as our partner in engineering, as we are part of the Copilot stack, we are going to keep working with engineers in Microsoft Research and Microsoft Engineering in India to push forward our product development around the world.

One common theme that we've been hearing from many technologists is how verbal skills and English skills will overtake math skills. How do you see this evolving?

I think we're going to see a democratisation of access to technology. Now here in India, many people speak fluent English better than most Europeans I'd say. But most kids still learn their mother tongue first - Hindi or in Bengaluru, Kannada.

Many of those dialects are already supported by the (AI) model today. So the universal programming language is human language. It has one big advantage. This is the way that we all learn to communicate from an early age.

So the fascinating thing is that you will be able to learn coding by just using the language that you're learning from your parents all over the world. But also, you will be able to express your ideas in human language, which is much more flexible than doing that in code. Then the model translates it into code - either behind the scenes or when you can still see the code and you can still understand the code.

India has a huge number of computer science graduates, we also produce some of the biggest number of engineers every year. What's going to happen to the future of coding. Will AI replace coders or engineers?

I don't see it replacing developers at all. It's just a natural evolution that software developers have gone through over the last five decades.

Number two is that this AI wave is actually generating new work. The demand for companies all over the world is to now build AI systems. So all these companies now also need AI experts and they're not finding them on the street. They're finding them by upskilling their internal developers.

People at GitHub and Microsoft that are building the copilots previously were called full-stack engineers. Now they're AI engineers and they're building these copilots. So the developers are moving up the stack and the complexity of the systems are constantly increasing. Most companies, if not all companies, are writing much more code every single day than they are deleting code.

So we're sitting on an ever -growing stack of software that exists in our worlds, including thousands, if not tens of thousands of open-source libraries, which is, by the way, another such trend where instead of building it all ourselves, we have used open-source within our stack, and that hasn't replaced developers either, it just created more work, ultimately, and helping us to move up the stack to build ever more complex systems. If you think about the challenge that we have in the world, there's a lot of software still to be built, to solve climate change, to solve cancer, diabetes. Those are all going to be AI systems supported by software.

Every company needs software developers, and soon enough, every company will need an AI engineer, a prompt engineer with these skills. So I'm very optimistic that we are still going to be in that place where being a software developer is a very attractive job category and I encourage everybody including my own kids to learn coding.

Read: Momentum around developers, development in India unbelievable: Microsoft CEO Satya Nadella

Has the bar significantly gone up?

The bar has massively gone up. There is a huge number of requirements on software developers to build international systems or global companies since they use thousands, if not tens of thousands, of servers spread around the world but have to comply with local regulations, local privacy protection, security.

There's a new model now almost weekly and so my advice to kids and parents is get started as early as possible, consider programming with the help of a copilot in the same way that you consider literacy, math, art and science and we want our kids to learn an instrument.

Learning prompting skills will be crucial for kids and even for professional workers as those prompting skills are universally translatable between the different copilots. If I know how to prompt, I can use that in ChatGPT to write a text. I can use it in Outlook to write an email, or in Powerpoint to create a deck. And I can use it in GitHub copilot to write an application.

Where is India in the global AI landscape? Are we missing out on building foundational models?

In a way, this is only the second year of AI post ChatGPT. A year ago, there wasn't much in the space of large language models. In India, if you look now, there's actually a bunch of projects happening.

I think there's an opportunity for India similar to how you see now with Mistral, for example, in Paris, or Stable Diffusion, that originally came out of a collaboration between different startups and research across Europe.

I don't think we can say that the ship has already sailed. It's early days. If we time travel back to 1994, the year after Netscape came out, we probably would have been very often predicting the journey that different companies and different countries have taken.

I think India is moving fast.

You've been integrating Indian languages like Hindi to GitHub Copilot. How has this helped in driving adoption?

The thing that's really important for us is that you can start coding and exploring these topics without having to learn English first, or without having to use English because you might not feel as confident in English that you are in Hindi or in Japanese or Brazilian Portuguese?

Number two is even in a professional workforce, many companies, when you move away from the international companies, are still communicating in the local language and we see that prevalent in all parts of the world. They don't want their internal communication to be in English. They want to be in Hindi, or they want to be in Japanese because that's part of their company culture and software, over the last decades, have kind of like forced companies to convert to English or have that hybrid world. But that makes work and processes harder.

Ultimately, the last thing is that we have legacy code, technical debt as developers would call it, in the world. Many companies still run Cobol on mainframes. It still runs in many financial services, local and federal government tax authorities in the world. It's code in a language that was invented in the '50s, written in the '60s and '70s.

That code is up to very different standards than what we call cloud native today, so many of those companies are interested in converting and modernising it. With Copilot, you are enabling people to understand them.