Open Source AI Presentation

This week, we had a presentation by Nick Vidal, who is from the OSI(Open Source Initiative). Nick introduced us to Open Source AI and what it entails. Mostly, we covered how open source AI is defined by OSI and projects that adhere to the definition, and also projects that appear to be open but actually go against some of the ideologies of open source.

Defining Open Source AI

Due to the nature of open source and AI, it is pretty difficult in creating a general consensus of what Open Source AI should be defined as. This is because AI is rapidly developing, and there will have to be new considerations on what counts as open or not. As Nick discussed, there are many companies that have AI technologies that are really not open even though they claim to be “open source”. Sometimes this includes obscuring source code or technologies that the project uses. The definition that we discussed for open source AI is very similar to the general open source definition in that it should be able to be used, studied, modified, and shared. Another interesting thing to note from the presentation is open source AI does not necessarily need to have its dataset be open, though it is preferred which was surprising to me.

The Future

AI is a very powerful tool and needs to be discussed in the context of open source since it will allow for greater technologies to be developed. If there is no foundation for how open source AI tools are developed, then it will be difficult to find ones that are actually what they claim to be. I think it is very important that these discussions are being had now, and hope that open source ideologies will be widely adopted by these tools so that newer and better technologies can be created.