FSF’s Pursuit of Freedom in Machine Learning Applications: A Late but Necessary Initiative

The Free Software Foundation (FSF) made a groundbreaking announcement on Tuesday, revealing their plans to focus on “freedom in machine learning applications”. This initiative goes beyond just the software itself, extending to the training data and associated scripts as well.

After months of deliberation, the FSF is nearing a conclusion on their official statement regarding free machine learning applications. According to their recent news release, all software included in a free ML application must provide users with the four freedoms that define free software. This includes not only the software processing training data but also the software interpreting model parameters for human-usable output.

The FSF believes that true freedom in machine learning applications requires that all training data and related scripts respect all users, following the four freedoms. This may even entail releasing model parameters for users to use and redistribute. Any ML applications that do not offer these freedoms are considered nonfree, regardless of the software components being free.

While the FSF’s efforts in the AI/ML space are commendable, some critics argue that they may be too little, too late. With the AI/ML industry already well-established, the FSF’s late entry into the conversation may limit their impact. However, for dedicated free software enthusiasts, the FSF’s commitment to ensuring freedom in machine learning applications is a step in the right direction.

As the FSF prepares to release their statement on free machine learning applications in late 2024, the tech community eagerly awaits to see the potential impact of their work. Will the FSF’s efforts lead to meaningful improvements in the realm of AI/ML software, or will they struggle to catch up to the industry’s rapid advancements? Only time will tell.