I personally have a strong connection to the open source community going back over 20 years. I was release manager for PHP, where I established the processes still largely in place today. Being a big believer in the value to society, open source provides. I will always keep an eye on what we can open source and want this to be part of the DNA of Witty Works.
However, Witty is not open source software. For now, I can say that we will always provide a freemium version. Our goal is to maximize the positive impact for society, and a freemium model is the most effective way to maximize our reach.
But what is Witty?
At its core, Witty is a system in the form of a so-called API (application programming interface) with a debiased language model that receives text, analysis it and provides feedback. However, a raw API is not going to make the impact we strive to make. What is needed are client programs that make it easy for users to send text to this API and to get the feedback presented in an easily consumable and actionable way. For this we have build the browser extension and we expect to build many more clients to cover other use cases, like Microsoft Word or Outlook, mobile keyboards, Slack integration etc.
These clients are an important part of the solution, even if they are essentially a distraction from our core mission: being able to give feedback on how to write more inclusively. As we would love it if we would not have to build those clients so that we can focus more of our energy on the core inclusive writing algorithm. As such, we also wouldn’t mind collaborating with other organizations who for whatever reason also require such clients. However, app store rules complicate this a bit, as they forbid submitting the same code for the same use cases. App store rules are also inherently arbitrary, especially if you are a young startup.
We actually hope that our browser extension will become obsolete as we are lobbying browsers and word processor providers to open up an API for their spell checkers to allow for a more fluent integration, which should also create a path towards optimal accessibility.
On the language model side we are working together with researchers at various universities and we hope through this path we will contribute to the community advancements in ethical AI with reduced bias and a better balance towards privacy. I expect this to be our most significant contribution for the open advancement of ethical tech.
Best practices for ethical AI startups
We are also in the process of establishing an AI ethics board. We started on researching this already this summer, but quickly realized there isn’t much in the way of shared experience out there on which we could base this effort. In medicine there is some inspiration we can draw upon but very little in the startup context. So this is another thing we want to do, share with the world not just the reports from our AI ethics board, but also how we created the board and what processes we have defined.
So in summary, at this point we are not sharing code with the world. However, we are collaborating with the research community. We are working towards more accessible access to browsers and word processors APIs. We are also working to establish best practices for ethical AI startups.
And of course, we are sharing our freemium solution to make sure anyone has the information to write inclusively.