GitHub CEO Thomas Dohmke is a busy man, not only found wandering the streets of Bengaluru, reminiscing old memories, but also captivating developers and enterprises alike in the city.
The company, which recently announced the technical preview of GitHub’s newest product- Copilot Workspace, showed live demos of how Copilot Workspace to developers in India.
Developers can interact with Copilot Workspace through chat as well as with a voice interface in multiple languages like Hindi, Kannada, and other Indian languages.
However, not everyone has access to this newest tool from GitHub. When AIM asked Dohmke if GitHub could announce the general availability of Copilot Workspace during their flagship event, GitHub Universe 2024, which is taking place in October, he said- no promises.
But when we posed the same question to Sharryn Napier, vice president of Asia Pacific, India, Japan & China, and GitHub, who accompanied Dohmke to India, her response was a little bit more optimistic. “Hopefully,” she said.
( Dohmke in Brigade Road, Bengaluru in 2008 and 2024)
GitHub is Wooing Customers in India
During his visit, Dohmke visited Infosys’ campus and launched the first GitHub Centre of Excellence in Bengaluru. At Universe 2023, held in San Francisco, Dohmke told AIM that Infosys was one of the early adopters of GitHub.
Moreover, Dohmke was also hobnobbing with executives from Paytm and Indian IT companies such as Tech Mahindra and HCLTech.
Some other notable customers for GitHub Copilot enterprise include LambdaTest, MakeMyTrip, Swiggy, Cognizant, Air India, Glance, and the Indian government-owned public portal Government e-Marketplace (GeM).
Given India’s ever-evolving tech landscape and the growing number of developers, it makes sense for GitHub to position India as an important market.
According to GitHub, over 15.4 million developers in India are building on GitHub, growing 33% year over year (YoY). Moreover, the Microsoft-owned company also revealed that India is set to overtake the US in terms of the number of developers by 2027.
Another interesting piece of data from that report is that Indian developers are the second-biggest contributors to generative AI projects on GitHub after the US.
Hence, it makes perfect sense for GitHub to become aggressive with sales activities in the country. Interestingly, GitHub also strongly positions itself in other Asian markets, such as Singapore.
( Dohmke at Github Constellation 2024)
Defining the ‘G’ in AGI
Given that GitHub is positioning AI as critical throughout the software development lifecycle, we asked Dohmke what his thoughts were on artificial general intelligence (AGI).
“Today I see no sign that machine learning models or Large Language Models (LLMs) have sentience. They are not creative. They are machines created by us that help us with the things that we want to do or don’t want to do,” he told AIM in an exclusive interaction.
However, this does not mean AI today can’t be transformative. Taking the example of code scanning, Dohmke said, “We shipped code scanning in public preview earlier this year that helps developers to fix their security vulnerabilities. So I think we are more focused on what we can do today with AI and how we can help developers write better codes without worrying much about the future.”
Nonetheless, Dohmke does make a very good point about AGI. He said everybody has a different understanding of what AGI even means and what the ‘G’ in AGI really stands for.
What Dohmke is indicating is how we even know when an AI system becomes conscious. Today, LLMs encompass all the knowledge available in the world, a trait unattainable by humans. Yet, does this superiority over human capabilities make them inherently better or even sentient?
Copilot Workspace vs Devin
GitHub’s attempt to woo developers and enterprises in India also makes sense, given that many alternatives to GitHub Copilot have emerged recently.
Most notably, Cognition Lab’s Devin, which the startup positioned as the world’s first ‘AI Engineer, ‘ is gaining prominence. Other tools, such as StarCoder 2 and Amazon Q for Developers ( previously Codewhisper), are also gaining prominence.
Previously, we have asked GitHub how Copilot Whisperer differs from Devin. Even though they are designed to solve similar problems, there are some fundamental differences, said Jonathan Carter, head of GitHub Next, in an earlier interaction before the technical preview announcement of Copilot Workspace.
“We don’t view GitHub Copilot Workplace as an ‘AI engineer’; we view it as an AI assistant to help developers be more productive and happier,” Carter said.
At GitHub Galaxy, we asked Dohmke his thoughts on Devin, and he gave a very CEO-esque answer. “First of all, competition is great. You don’t want to play a sport without any competitor, so we embrace competition.”
But he did share his thoughts on what makes Copilot different from Devin. The difference is the approach. While others [Devin] have taken an autonomous first approach, Copilot Workspace helps the developers through the software development life cycle. I think both of these approaches are valid and they’re probably just on different timescales when they’re leading to success,” Dohmke said.
AI is not taking over jobs, but it is redefining job roles?
At Galaxy, most of the questions directed towards Dohmke were concerned with AI’s impact on jobs. Dohmke’s response to those questions was reassuring.
“In many ways this new age of AI has actually created more demand on developers because now somebody has also to build all the AI systems,,” he said.
The recent boom in AI is creating newer AI companies in Silicon Valley and, likewise, in India. In the last year or so, many generative AI companies have mushroomed in India.
“Humanity has so many problems to solve and there is so much work to be done. We are not going to run out of work.”
However, many experts AIM has spoken to earlier are of the opinion that AI will be able to do what an entry level coder is required to do in an organisation. However, Dohmke had a different opinion.
“When you embark on a new job, whether fresh out of college or transitioning from another company, the primary challenge is understanding the company’s operations and code bases, which could be thousands of files.”
A tool like Copilot could be very useful for entry-level coders. Moreover, Dohmke believes that going forward, applicants will be expected to be adept at leveraging AI tools like Copilot and ChatGPT.
“Some software companies have even begun incorporating Copilot into their interview processes, replacing traditional coding exercises with tasks that assess applicants’ ability to utilise these tools effectively.
“This shift underscores the evolving skill sets required in the workforce, emphasising adaptability and proficiency in leveraging AI technologies over rote memorisation of coding syntax.”
Moreover, Dohmke also advocates for teaching students in colleges the integration of AI coding assistants.
“I believe graduates will emerge from university with a much stronger skill set. We’ve already observed this trend in colleges transitioning from teaching Python to Rust, a more intricate programming language. This shift is driven by the substantial opportunities it presents for developers, particularly those proficient in Python. I anticipate witnessing more such transitions in educational institutions.”
Insure codes?
According to GitHub, using AI makes coders 55% faster than those who don’t use any AI coding assistant. GitHub Copilot has also become the most widely used AI tool by developers.
However, concerns have been raised about the efficiency of the codes generated by these tools. Findings from the code reviewing tool GitClear indicate that the proliferation of AI coding assistants in 2022 and 2023 coincided with an increase in the frequency of code requiring correction within two weeks of its creation.
Dohmke shared an anecdote to respond to this. “I spoke with a CEO who was coding with his son. They needed to convert binary to decimal and initially wrote 150 lines of code. Upon consulting Copilot, it condensed the solution to just two lines by leveraging an open-source library.”
He pointed out that this demonstrates increased efficiency in code management, a crucial aspect for companies. “Moreover, if you don’t want to use that open-source library, it still gives you code without the open-source library.”
The trick, according to Dohmke, is in prompting.
“While the best programmers may outperform the model, the focus in larger companies is on raising the average skill level. Studies also show improvements in pull request and merge rates, indicating enhanced developer focus and quality during the development process,” he added.
