Five Ways Open Source Software Is Shaping AI Politics


Open source software (OSS), which is free to access, use and modify without restrictions, plays a central role in the development and use of artificial intelligence (AI). An AI algorithm can be thought of as a set of instructions, i.e. what calculations should be done and in what order; developers then write software that contains these conceptual instructions as actual code. If this software is then released in an open source manner, where the underlying code is publicly available for anyone to use and modify, any data scientist can quickly use this algorithm with little effort. There are thousands of AI algorithm implementations that make it easier to use AI in this way, as well as a critical family of emerging tools that enable more ethical AI. At the same time, there is a decreasing number of OSS tools in the particularly important sub-domain of deep learning, which strengthens the influence in the market of companies that develop this OSS, Facebook and Google. Few of the AI ​​governance documents focus enough on the role of free software, which is a regrettable oversight, although it quietly affects nearly every issue in AI politics. From research to ethics, and competition to innovation, open source code plays a central role in AI and deserves more attention from policy makers.

1. OSS accelerates adoption of AI

OSS enables and increases the adoption of AI by reducing the level of mathematical and technical knowledge needed to use AI. Writing the complex math of algorithms into code is difficult and time consuming, which means that any open source alternative that exists can be of huge benefit to data scientists. OSS benefits from both a collaborative and competitive environment as developers work together to find bugs as often as they compete to write the best version of an algorithm. This often results in more accessible, robust, and high-quality code compared to what an average data scientist, often more of a data explorer and pragmatic problem solver than a pure mathematician, might develop. This means that well-written open source AI code dramatically increases the capacity of the average data scientist, enabling them to use more modern algorithms and machine learning features. So while much attention has been paid to training and retaining AI talent, making AI easier to use, as OSS code does, can have an equally large impact in enabling the economic growth of AI.

2. OSS helps combat AI biases

Open source AI tools can also enable wider and better use of ethical AI. Open source tools like IBM’s AI Fairness 360, Microsoft’s Fairlearn, and University of Chicago’s Aequitas alleviate technical barriers to tackling AI bias. There is also OSS software that makes it easier for data scientists to query their models, such as IBM’s AI Explainability 360 or Chris Molnar’s interpretable machine learning tool and book. These tools can help time-pressed data scientists who want to build more responsible AI systems, but who are under pressure to complete projects and deliver for clients. While increased government oversight of AI is certainly needed, policymakers should also consider investing in open source ethical AI software more frequently as an alternative lever to improve AI’s role in society. The National Science Foundation is already funding research into the equity of AI, but funders and foundations should see free software as an integral part of ethical AI, and further fund its development and adoption.

3. AI OSS tools advance science

In 2007, a group of researchers argued that “the lack of freely available algorithmic implementations is a major obstacle to scientific progress” in an article titled “The Need for Open Source Software in Machine Learning”. It’s hard to imagine this problem today, as there is now a plethora of AI OSS tools for scientific discovery. As an example, the open source AI software Keras is being used identify the subcomponents of mRNA molecules and build neural interfaces to better help blind people see. OSS software also makes it easy to replicate research, allowing scientists to verify and confirm each other’s results. Even small changes in the way an AI algorithm has been implemented can lead to very different results; the use of shared free software can mitigate this source of uncertainty. This makes it easier for scientists to assess the research results of their colleagues, a common challenge across many disciplines facing an ongoing replication crisis.

While OSS code is much more common today, efforts are still being made to increase the percentage of academic papers that publish their code, currently around 50-70% at major machine learning conferences. Policy makers also have a role to play in supporting the OSS code in science, for example by encouraging federally funded AI research projects to publish the resulting code. Granting agencies may also consider funding the ongoing maintenance of AI OSS tools, which is often a challenge for mission-critical software. The Chan Zuckerberg initiative, which funds critical OSS projects, writes that OSS “is crucial for modern scientific research … yet even the most widely used research software lacks dedicated funding.”

4. Free software can help or hinder competition in the tech industry

The OSS has important ramifications for competition policy. On the one hand, the public release of the machine learning code broadens and better enables its use. In many industries, this will enable greater adoption of AI with less AI talent, possibly a clear advantage for the competition. However, for Google and Facebook, the open source of their deep learning tools (Tensorflow and PyTorch respectively) could further strengthen them in their already strengthened positions. Almost all of the developers in Tensorflow and PyTorch are employed by Google and Facebook, which suggests that companies aren’t giving up a lot of control. While these tools are certainly more accessible to the public, the oft-stated goal of “democratizing” technology through free software is, in this case, an understatement.

Tensorflow and PyTorch have become the most common deep learning tools in industry and academia, providing great benefits to their parent companies. Google and Facebook benefit more immediately from research conducted with their tools because it is not necessary to translate academic findings into a different language or framework. Additionally, their dominance manifests a pipeline of data scientists and machine learning engineers trained in their systems and helps them position themselves as the cutting edge companies to work for. Overall, the benefits for Google and Facebook of controlling OSS deep learning are significant and may persist into the future. This should be taken into account in all discussions on competition in the technology sector.

5. OSS creates AI standards by default

OSS AI also has important implications for standards bodies, such as IEEE, ISO / JTC and CEN-CENELEC, which seek to influence AI industry and policy. In other sectors, standards bodies often add value by disseminating best practices and enabling interoperable technology. However, in AI, the diverse use of operating systems, programming languages, and tools means that interoperability issues have already received considerable attention. Additionally, the AI ​​community of practitioners is somewhat informal, with many practices and standards disseminated via Twitter, blog posts, and OSS documentation. The dominance of Tensorflow and PyTorch in the deep learning subdomain means that Google and Facebook have disproportionate influence, that they may be reluctant to give in to consensus-driven standards bodies. So far, OSS developers have not been widely involved in the work of international standards bodies, which can significantly inhibit their influence in the field of AI.

AI policy is linked to open source software

From research to ethics, from competition to innovation, open source code plays a central role in the development of the uses of artificial intelligence. This makes the constant absence of open source developers in policy discussions quite notable, as they exert significant influence over the direction of AI and very specific knowledge of it. Involving more AI OSS developers can help AI policymakers more systematically consider the influence of OSS in furthering the fair and equitable development of AI.

The National Science Foundation, Facebook, Google, Microsoft and IBM are donors to the Brookings Institution. The results, interpretations and conclusions published in this article are solely those of the authors and are not influenced by any donation.

Source link


Leave A Reply