Mozilla is going to work on ‘reliable’ open source AI with new start-up
Mozilla creates a start-up called Mozilla.ai. With this, the company wants to work on ‘reliable’ and open source AI. Mozilla is investing $30 million in the initiative, which will be led by AI researcher Moez Draief.
Mozilla announces the creation of Mozilla.ai in a blog post and says it is working on an ‘ecosystem’ of reliable and open source AI. The company will initially ‘create tools that make generative AI safer and more transparent’. Mozilla.ai also says it’s working on “people-centered recommendation systems that don’t misinform us or undermine our well-being.” The start-up shares little to no concrete details, but says it will provide more information in the coming months.
The Mozilla Foundation is initially investing $30 million in the new company. Moez Draief becomes the managing director of Mozilla.ai. Draief was previously the head of Huawei’s AI lab and global chief scientist at consulting firm Capgemini. Mozilla’s executive director Mark Surman, Karim Lakhani of Harvard University and Navrina Singh of AI company Credo will become the first board members.
The full team initially consists of 25 engineers, scientists and product managers, says Mozilla in an interview with TechCrunch. The company wants to work with that team on reliable recommendation systems and large language models along the lines of OpenAI’s GPT-4. In time, however, the start-up wants to build a network of “affiliated companies and research groups that share Mozilla’s vision.”
Mozilla has been working on AI for some time. The company published in 2020 a white paper with his vision on reliable AI models. However, Mozilla has not released concrete AI-based products yet. Now that AI is on the rise with models such as OpenAI’s GPT-4 and ChatGPT, Google Bard, Stable Diffusion and Dall-E, Mozilla says it wants to work on models that are more open and transparent. “We’re not just wondering what’s possible and how people can benefit from it. We’re also wondering what could go wrong and how we can handle it,” the company writes.