Chanticleer
The CEO of Microsoft-owned Github says the sheer complexity of a world dominated by software code means AI is needed to help fill the tech skills gap.
Every now and again, Thomas Dohmke likes to print out his boarding pass. In a strange way, that little slip of paper helps remind the chief executive of the Microsoft-owned GitHub just how much of the rest of his life is now reliant on software.
And as Dohmke knows only too well, a good chunk of the software used in our daily lives has been developed on GitHub’s web-based platform, which allows multiple developers to collaborate on the same project using an open-source tool called Git. A staggering 94 million developers are part of the GitHub community.
Thomas Dohmke says AI can make developers more productive – and happier.
Dohmke, who grew up in Germany and fondly remembers when his Wednesday night computer club meeting was one of the few ways a budding developer could get outside help, remains a passionate advocate for open-source software, which he argues means “people around the world can collaborate in what I’d like to call the largest team sport on Earth”.
But he says a world that is so reliant on software is also a world that is incredibly difficult for developers to manage.
“Software has eaten the world. And the way it has happened, and will continue to happen, has increased the complexity by an order of magnitude for many developers,” Dohmke tells this column ahead of a visit to Australia this week.
“We have arrived in a world where we have so much code, and we have such a high complexity, that we need another approach, we need a new approach to allow us to go the next step on the ladder.”
That next step, Dohmke believes, is artificial intelligence.
When Microsoft acquired GitHub in October 2018 for $US7.5 billion ($9.8 billion at the time), there was great concern in the developer community that the deal would damage the deep sense of community that had flourished across the code sharing platform.
That user numbers have trebled since then would suggest those fears have proven to be unfounded. Dohmke, who sold his own business HockeyApp to Microsoft in 2014 and then led the tech giant’s team that worked on the GitHub acquisition, says this is down to the fact Microsoft fully embraced the principle that GitHub would always put developers at the centre of what it does.
While GitHub’s business has expanded – it makes money by offering a suite of software development and collaboration tools to enterprise customers, and recently hit $US1 billion in annual recurring revenue – Dohmke, who succeeded Nat Friedman as the company’s CEO in November last year, says this developer-first mentality hasn’t changed.
While GitHub’s valuation is tiny in the context of Microsoft’s $US1.8 trillion market value, comparisons with listed peers including GitLab and Australian-born Atlassian (which trade on 15 times sales and 11 times sales respectively) suggest a rough, back-of-the-envelope valuation of between $US11 billion and $US15 billion for GitHub.
But if Microsoft has largely left GitHub to get on with what it does best, the deal has also allowed GitHub to tap the extraordinary resources of the mothership.
One way is via Microsoft’s powerhouse Azure cloud business; GitHub can piggyback off the huge network of data centres to improve the performance of its own platform.
But perhaps the most powerful example of collaboration is in artificial intelligence, where GitHub has been able to work with Microsoft and research laboratory OpenAI to launch a product called Copilot.
(OpenAI is also the firm behind ChatGPT, the conversational chatbot that has become the talk of the tech world for its ability to have a crack at everything from answering trivia questions and summarising meeting notes to sorting complex text files and – ahem– writing articles.)
Users of Google’s Gmail platform, or its Google Docs product, will likely be familiar with the simple (but eerily effective) AI technology that predicts text as the user types – offering a greyed-out suggestion to complete a phrase or sentence that the user accepts by hitting the “tab” button or swiping right on their mobile.
Copilot uses the same principle, but for developers. As the developer writes code, Copilot will throw up a suggestion for the next line. And Dohmke says given code is typically written in a formulaic way, the suggestions are arguably more helpful than in an email or a document, where the complexity of language is greater.
The AI is trained off GitHub’s broad data set, but also adapts to the user’s mannerisms and approach. “Your Copilot works differently than mine, and becomes this kind of a personal assistant that is fully adapted to your style,” Dohmke says. “But developers are still the ones that have to prompt the AI, they are the creator.”
Since its launch in October 2021, Dohmke has seen some impressive results from developers who’ve taken a punt on Copilot.
Where the AI system is engaged, Copilot is writing about 40 per cent of a developer’s code, which Dohmke says is about double his expectations. But his favourite test is where GitHub asked two groups of developers to take on the routine task of building a web server. The group using Copilot got it done in 71 minutes, while the group working alone needed 148 minutes.
“It’s just paradigm shifting. It shows you how powerful this can be and how much it is going to change our professional lives,” he says.
Making developers happier is a big part of Dohmke’s pitch. That doesn’t exactly sound business-minded, but there is some science to it.
With the demands on developers increasing, there is real value to keeping them in what Dohmke calls “the flow state” so they don’t have to stop coding to do something else, such as researching a bit of code that Copilot could provide.
“Many creative people, as long as they are in the flow state, they’re actually able to create amazing things, and have this high productivity level that isn’t actually perceived as hard work, it’s perceived as enjoyment. You get dopamine hits, you get adrenaline, you get this feeling of, ‘oh, this is amazing’.”
Dohmke says the next few years are about refining and improving Copilot: more powerful hardware, more data to train the system on and different user interfaces for developers. There’s an education effort required too, given the way developers tend to work.
The power of this really is that it is kind of like a second brain that you get as a human.
— Github CEO Thomas Dohmke
“They are often quick to be sceptical of changes – never change a running system is a slogan from the developer world,” Dohmke says.
Still, he sees the world as standing at the beginning of an important era of 10 years or longer where AI will make giant evolutionary strides. But that rate of development is going to rely largely on people, and tech skills remain in short supply, particularly in the Western world.
“Even with the tech layoffs, I believe we do not have not enough developers to solve all the challenges of the future, where software is at the core,” Dohmke says, giving the example of self-driving cars, where the level of technical challenge is incredibly high.
While the GitHub boss sees the need to train more developers and software engineers, he argues AI will have to step up to help fill the tech skills gap.
“I think AI is going to drive the next wave of innovation across not only the software industry, but really across all the industries,” he says.
“We truly believe that you will have a Copilot for everything. Whether you’re a car mechanic or a doctor, you can kind of already imagine how that would run. The power of this really is that it is kind of like a second brain that you get as a human.”
Whether you think that sounds exciting, slightly scary or both, one thing is certain – it’s going to require a hell of a lot of code.
Follow the topics, people and companies that matter to you.
Fetching latest articles
The Daily Habit of Successful People
Leave a Reply