Ubisoft Teams With Mozilla to Develop New Coding AI Technology
The Clever-Commit tech was developed by the Ubisoft Technology Group to "help programmers evaluate if a code-change will introduce a new bug by learning from past bugs and fixes."
Ubisoft has teamed with Mozilla, makers of the Firefox web browser, to develop and apply a new AI technology called the Clever-Commit to assist in coding by identifying and fixing bugs.
The tech was developed by the Ubisoft Technology Group to "help programmers evaluate if a code-change will introduce a new bug by learning from past bugs and fixes."
Yves Jacquier, head of Ubisoft's research lab, La Forge, revealed the partnership during his talk at this year's D.I.C.E. Summit in Las Vegas.
Clever-Commit was first introduced last year as a prototype named Commit-Assistant.
"AI has been using games for years," Jacquier tells The Hollywood Reporter, referencing the 1996 chess matches held between IBM's supercomputer Deep Blue and chess champion Garry Kasparov.
The new tech will be used to help Ubisoft roll out higher quality games at a faster pace and has already been used in the development of major Ubisoft games with plans to integrate the tech into other brands in the future. Meanwhile, Mozilla will use the tech in its development workflow.
"What is new is that both AI and video games have reached a new level in the past three or four years," Jacquier says. "Open worlds are more and more like rich simulations of the real world. You have a new AI called deep learning, which is all about using complex data, extracting some features from it to be able to make predictions when it sees new data."
Jacquier sees games and the new tech as a potential testing ground for real-world tech such as autonomous self-driving cars. "This new relationship between AI and video games is now starting to have positive impacts in transportation, health care, education," Jacquier says.
With AI developed to handle some of the more mundane tasks of game development and coding, Jacquier says that creators can now focus on more "exciting things."
"Machine learning is good at generalizing and automating the most redundant tasks," he says. "When a programmer creates a bug, someone will test that, file that, correct that, retest that — that's a lot of work. What if we used this energy to focus not on solving bugs but focusing on new, more exciting features? That's really the idea of what we're doing."