On the eve of its 25th anniversary, Mozilla, the nonprofit behind the Firefox browser, is launching an AI-focused startup.

called Mozilla.aiThe new company’s mission isn’t to build just any AI — its mission is to build AI that is open source and “trustworthy,” said Mark Surman, Mozilla’s executive president and head of Mozilla.ai.

“I’ve been working on reliable AI for almost five years and I’ve felt a constant mix of excitement and fear,” he told australiabusinessblog.com in an email interview. “The past two months of lightning fast big tech AI announcements have been no different. There’s some really exciting new technology emerging – new tools that have instantly empowered artists, founders…all sorts of people to do new things. The fear comes when you realize that hardly anyone is looking at the guardrails.”

Surman referred to the proliferation of AI models in recent months that, while impressive in their capabilities, have worrying implications for the real world. At release, OpenAI’s text-generating ChatGPT could be asked to write malware, identify exploits in open source code, and create phishing websites that resembled high-traffic sites. Text-to-image AI like Stable Diffusion, meanwhile, has been co-opted to create pornographic, non-consensual deepfakes and ultragraphic depictions of violence.

The makers of these models say they are taking steps to prevent abuse. But Mozilla felt that not enough was being done.

“We have been working on reliable AI in public interest research for about five years, hoping that other players in the industry with more AI expertise would step up to build more reliable technology,” said Surman. ‘They don’t. So we decided in the middle of last year that we had to do it ourselves – and find like-minded partners to do it with us. Then we started looking for someone with the right mix of academic and industrial AI experience to lead it.”

Funded by a $30 million seed investment from the Mozilla Foundation, Mozilla’s parent organization, Mozilla.ai is a wholly owned subsidiary of the Mozilla Foundation — just like the Mozilla Corporation (the organization responsible for developing Firefox) and Mozilla Ventures ( the VC of the Mozilla Foundation). fund). Its general manager is Moez Draief, who was previously chief scientist at Huawei’s Noah’s Ark AI lab and the global chief scientist at consultancy Capgemini.

Karim Lakhani of Harvard, Navrina Singh and Surman of Credo will serve as Mozilla.ai’s first board members. Lakhani is the chairman and co-founder of Harvard’s Digital, Data and Design Institute, while Singh serves on the US Department of Commerce’s National AI Advisory Committee, which advises the president on a range of ethical AI issues.

Surman describes Mozilla.ai as part research company, part community – a startup dedicated to helping create a trusted, independent open source AI stack. Initially, Mozilla.ai’s priority will be to build a team of about 25 engineers, scientists, and product managers to work on “trustworthy” recommendation systems and large language models along the lines of OpenAI’s GPT-4. But the company’s broader ambition is to build a network of affiliates and research groups — including Mozilla Ventures-backed startups and academic institutions — that share its vision.

“We think there is a commercial market for reliable AI – and that market needs to grow if we are to change how the industry is embedding AI into the apps, products and services we all use every day,” said Surman. “Mozilla.ai — working loosely with many affiliates, researchers, and governments — [has] the opportunity to jointly create a ‘trust first’ open source AI stack. If we are successful, the industry mainstream could draw from this stack as part of their mainstream toolkit, just as they have done with the Linux and Apache stack for the past two decades.”

Mozilla.ai won’t do it alone – not quite. Several nonprofits are on a mission to democratize AI tools, including the recently founded EleutherAI Institute, funded by corporate donors including Canva and Hugging Face. There’s also the Allen Institute for AI, founded by the late Microsoft co-founder Paul Allen, and the Alan Turing Institute. Smaller promising efforts include AI startup Cohere’s Cohere For AI and Timnit Gebru’s Distributed AI Research, a global decentralized research organization.

It is significant that Mozilla.ai is not a non-profit organization. Although it is bound by certain ethical principles (namely the Mozilla Manifesto), it is open to spinout – and indeed wants to spinout – its more successful explorations into products and companies in addition to open source projects.

Draief sees this as an asset rather than a drawback, arguing that it gives Mozilla.ai the flexibility that nonprofits lack. According to him, there are cautionary tales such as OpenAI, which was founded in 2015 as a non-profit organization but was later forced to move to a “capped profit” structure to fund its ongoing research.

“The big question is, how many of the newer, smaller reliable AI startups will be able to remain independent?” Draief told australiabusinessblog.com via email. “It is clear that the big players – especially the Amazon, Google and Microsoft cloud platforms – are rushing to consolidate the AI ​​space. This is where all the money is made. And it will be hard for small businesses not to get sucked into this consolidation.”

On the hunt for current AI research trends — and, not coincidentally, the better-funded areas of research — Mozilla.ai will be developing tools over the coming months that will allow users, for example, to query the sources behind the answers AI chatbots give them. The company will also look to create systems that give users more control over AI for content recommendation (i.e. the algorithms that power YouTube, Twitter and TikTok feeds), such as systems that optimize a recommendation for individual or community values ​​- building on Mozilla’s existing research.

Draief isn’t arguing that shifting the AI ​​stack in a meaningful way will be a quick process. While he promises that Mozilla.ai will ship code “this year,” he’s speaking in terms of several years.

But measurable success takes more than time.

If history is any indication, voluntary frameworks and one-time tools won’t move the needle much, if at all. Mozilla.ai’s challenge will be to convince the industry that its vision of trustworthy AI is the right one — and to adopt that vision.

“Reliable AI features like this feel like they should be trivial to add, but we still see them mostly in the lab,” Draief said. “Mozilla.ai will work with researchers to turn their work into working code and make it possible to use it alongside more traditional AI tools.”