In an effort to promote the ethical and socially acceptable use of generative artificial intelligence and wide language models (LLMs), Mozilla organized the “Responsible AI Challenge.” This challenge was aimed at driving trusted AI solutions that served the whole of society rather than exclusively benefiting greedy mega-corporations or extremely wealthy individuals. Now, after an exhaustive review process, the Firefox Foundation has announced the winners of the challenge’s top prizes, who will receive funding to bring their projects to market.
Sanative AI takes top prize with its innovative anti-AI watermark technology
After careful deliberation by a judging panel comprised of AI academics, developers, and entrepreneurs, Mozilla has chosen Sanative AI as the first prize winner, awarding them a $50,000 bounty. This company is developing a revolutionary solution called “anti-AI watermark”, designed to protect original images and artwork from being misused as training data for generative AI algorithms and diffusion models.
Sanative AI’s adversarial AI tools technology
Sanative AI has created adversarial AI tools that act as safeguards against exploitation by AI applications. Its first product consists of a technology that applies a “small” visual noise to images uploaded to the Internet. This noise has minimal or even invisible visual impact on the mentioned images, but it confuses machine learning algorithms, resulting in “distorted” results.
Kwanele Chat Bot and Nolano are also recognized for their responsible AI solutions
Second prize in the Mozilla Responsible AI Challenge, with a reward of $30,000, went to Kwanele Chat Bot. This technology has been designed to protect women living in violent communities by giving them quick access to help and ensuring the collection of “admissible” evidence. This innovative solution seeks to provide support and security to women in dangerous situations.
On the other hand, the third prize, endowed with $20,000, was awarded to Nolano, a newly trained language model that uses natural language processing to run LLMs on local (offline) machines such as laptops or smartphones. This proposal focuses on allowing the use of wide language models without depending on an internet connection, thus providing greater accessibility and autonomy to users.
Additional support and opportunities for Responsible AI Challenge winners
In addition to cash prizes, the three winners of the Responsible AI Challenge will receive mentorship from AI and technology industry leaders. They will also have access to Mozilla resources and communities as they continue to develop and refine their projects. Britney Crooks, Mozilla’s Director of Product Strategy and Innovation, said the evaluation process was designed to be motivating for all participants, inspiring them to continue developing responsible AI solutions even if they weren’t winners.
Source: Mozilla Blog
