The CEO of the Github coding platform is a programmer himself.
Redmond “We are facing a revolution,” says Thomas Dohmke. As the CEO of Github, he is the overseer of a sort of gigantic library of computer code. Github acts as a platform where programmers can store, manage, and openly discuss different versions of their software. Dohmke is now introducing the next stage of the platform and gave the Handelsblatt a preview of the new technologies.
With the Copilot X project, the 44-year-old has the knowledge on the platform evaluated and analyzed by an artificial intelligence (AI) and then delivered in a way that is suitable for programmers. In practice, the offer should act like an AI assistant that supports programmers. He should not only find errors or solve problems, but also be able to write entire programs himself.
The IT expert is convinced: “This is a powerful tool.” When he founded his first start-up ToDoSoft at the end of the 1990s, he would have liked to have had such support. “We sent our software on cassettes,” recalls Dohmke.
Github: Artificial intelligence should help programmers
Together with his brother, he built up a company, ToDoSoft, which wrote computer programs for insurance companies. This enabled brokers, for example, to calculate suitable conditions for their customers.
Dohmke studied technical computer science at the Technical University of Berlin and then did his doctorate. He also built other companies. His company hockey app was finally bought by Microsoft at the end of 2014 and Dohmke moved to the Microsoft headquarters in Redmond.
Copilot X is software giant Microsoft’s latest push to integrate AI into all of its products. Microsoft bought Github in 2018. Dohmke has been CEO of Github since November 2021 and is one of the most influential IT experts from Germany in the US technology scene. Microsoft boss Satya Nadella had announced: “Artificial intelligence will penetrate all areas.”
>> Read also: How ChatGPT and Sam Altman will impact your life
Last week Nadella presented an extended version of Office programs such as Word, Excel and Powerpoint, in which an AI assistant is integrated. They can write texts, evaluate tables or design a PowerPoint presentation with suitable photos and animations using just a few keywords.
Copilot X should work similarly. In a chat window, programmers should let the AI assistant take over tasks. At the same time, checking and managing the code should become easier.
Microsoft: Start-up OpenAI attracts skilled workers
Behind the assistant is a partnership between Microsoft and the start-up OpenAI. OpenAI provides the language model that powers Copilot X. According to industry estimates, Microsoft is said to have invested around 13 billion dollars in the young company.
100
Github has millions of users on the developer platform.
But why does a software giant like Microsoft need help from a small start-up like OpenAI in the first place? “Excellent specialists would rather work for a company like OpenAI than for a large corporation like Microsoft,” says Dohmke. Microsoft didn’t want to compete with OpenAI, but was looking for a partnership.
Copilot X is not the first application in which Dohmke uses the capabilities of OpenAI within Github. For the previous version, which is simply called Copilot, an OpenAI model is also used. Github has more than 100 million users on the developer platform. For the fee-based additional service Copilot, Github has more than 4000 corporate customers. The copilot usually costs around ten US dollars per month.
The software already enjoys a great reputation in Silicon Valley. Cloud provider Box CEO Aaron Levie says Copilot is a great help with coding. The software acts as if a brilliant programmer is constantly looking over your shoulder while writing code. In addition, Copilot could help to create software in programming languages that you do not know yourself.
>> Read also: This is how Microsoft and Google are creating a new reality
Nevertheless, even the improved version of Copilot is not without bugs. “In the end, someone has to be in control,” explains Dohmke. The OpenAI models repeatedly produced false statements in tests by various experts.
However, Dohmke softens the criticism: “No one is infallible.” This also applies to artificial intelligence. That is why it is so important that experts check the results of the AI assistant again at the end.
More: Microsoft integrates AI Copilot into Office