Open Source Speakers

This week, we had Nick Vidal from the Open Source Initiative give a lecture on open source AI and the mission of the OSI.

Nick’s presentation gave a great look into what it takes to uphold the meaning of “open source” as technology continues to evolve. He primarily spoke about AI and the OSI Open Source AI Definition, which was released recently. AI has exploded over the past three or so years and has caused a lot of discourse surrounding what open source means in an AI context, so Nick’s knowledge was very helpful. Companies like OpenAI that dominate the market are largely private (despite OpenAI’s name), and projects that claim to be open source (DeepSeek, LLama, etc) are often murky in how they fit into the definition of open source. Because of this, having a clearly outlined definition like OSI is providing is key to being able to understand which projects are truly open source and which are using the term to generate buzz. One of the most interesting points of Nick’s talk was about open sourcing training data, which can obviously be shaky ground if the model is trained on confidential information like medical data or academic records. As Nick said though, data for open source AI can be classified in several categories, some of which do not require the data to be pubicly accessable for the project to be defined as open source. In my opinion, open source AI will be an essential part of making progress in the field and loosening the grip of companies like OpenAI, so Nick’s talk was very enlightening.

We were also scheduled to hear Olivier Poupeney from FINOS speak, but it was unfortunately canceled. However, reading about FINOS was very interesting. We often think of private industry and open source as operating at odds or at least separately. It was therefore interesting to see how many financial companies use and contribute to open source projects. It would have been great to hear Olivier go deeper into the subject (I specifically wanted to hear more about how security is handled when developing open source projects that deal with such sensitive data), but it was still insightful reading about FINOS