AI for Good Global Summit Day 2: AI for the Masses: Democratizing AI to power social change
Jorren is a data scientist at Xomnia and runs our SustAIn program. His background is Artificial Intelligence with a specialism in Robotics. He recently traveled to the AI for Good Global Summit in Geneva and is sharing his favorite lessons learned.
Scaling AI for Good, or any AI solution for that matter, requires several ingredients: problem owners, data, AI, and a platform. Depending on who you are, you probably already have one of these things, but in many cases the other ingredients are distributed over a host of other parties. For example, at Schiphol Airport in Amsterdam, Operations has a problem with on-time performance that they want to solve with AI, but Digital has the expertise and the platform, and the ground handlers and airlines have the data. With AI for Good, the stakeholders are often more diverse. Take sustainable trade initiatives like certified sustainable lumber – Corporations often have the data on land-use, sustainable certification NGOs have a platform and want to solve the problem, and the AI is missing entirely from the equation.
There are a couple projects working on tying these pieces together:
AI requires huge data-sets, entrenching capabilities with large companies like Google and Facebook. The Ocean Protocol aims to be an ecosystem for the data economy, with a tokenized service layer that securely exposes data, storage, compute and algorithms for consumption. It’s a way to foster access to large relevant data sets for small companies and nonprofits. They are able to build trust in data-sharing through the use of blockchain – owners retain control over their data, and purchasers are able to access and combine valuable datasets. The initiative makes AI accessible to all.
But what if you have only problems or ideas, and no data or AI skills? OpenML is a platform that offers access to thousands of open datasets, and allows your to easily transfer data to a Machine Learning environment where you can build and train your models, and collaborate with others. The democratization of machine learning is crucial to empowering small organizations, communities, and nonprofits to grow their impact.
If you have a vision, but no data or platform, take some inspiration from Wildbook. They’ve created an AI recognition system that is building them a dataset by crawling YouTube. The system uses footage of animals in holiday videos to identify individual animals. The animals are tracked by location over time and the data is used to do population counts and to track migration patterns. They provide an example of leveraging existing technology to develop new, deep understanding.
If there’s one lesson to take on growing AI for Good, it’s that scaling must be a multistakeholder endeavor. We’ve intuitively grasped this at Xomnia through our yearly Datathons, which we always do with one or more partners who provide data, problems, or both. Two years ago we partnered with the police and Stop Pesten Nu, and last year with the World Bank. You can expect to see more of these multi-stakeholder initiatives from us soon.