The Open-Source Conundrum in AI

Though open-source software (OSS) affects nearly every façade of AI, it is largely absent from discussions around AI policy, according to a report by The Brookings Institute. The reason? Policymakers have barely noticed its influence in the rapidly evolving field of AI.

The heart of the matter revolves around the role of AI in our lives today amid growing concerns about transparency and fairness. There is hence a need to defend against the insidious impact of bias and ensure that crucial decisions are not inadvertently skewed by it.

But even as AI faces greater scrutiny, either in the form of private and public sector agencies enacting AI frameworks or by regulatory bodies seeking to police the development and direction of AI, the outsized role of OSS on AI development is overlooked, writes Alex Engler, author of the report and a fellow in governance studies at Brookings.

One example cited by the report was how the recently proposed European AI regulation does not address the role of OSS at all.

OSS and AI

The impact of OSS on AI is undeniable, argues Engler. OSS AI platforms and toolkits significantly lower the technical barriers for individual data scientists to leverage AI. This is especially true as not every data scientist might have the necessary training or mathematical background to implement complex algorithms for machine learning or deep learning.

Given that the most advanced AI toolkits are largely free and publicly available, OSS is quietly and inexorably affecting nearly every aspect of AI. And though OSS AI tools are enabling the faster adoption of AI in science and industry, those involved in its creation are not actively engaged or consulted in establishing AI standards.

“[The] data science community is somewhat informal, with many practices and standards disseminated through Twitter, blog posts, and OSS documentation… it is not clear that OSS developers are extensively involved in the ongoing AI standards discussions [with standards bodies],” noted Engler.

Finally, the current dominance of top machine learning platforms such as Tensorflow and PyTorch means that private firms such as Google and Facebook have outsized influence in the development and common use of deep learning methods – and should be taken into consideration by policymakers.

For now, Engler suggests involving more OSS AI developers to help AI policymakers better consider the influence of OSS as they seek a future of just and equitable AI development.

Image credit: iStockphoto/Visual Generation