Causal AI for Decision Making

Key Takeaways

I'm Chung Zhang from Microsoft research Cambridge I'm a principal researcher in machine learning today I'm going to talk about our project Corsica so this is a project about decision optimization with causal machine learning so the Holy Grail notification optimization is the ability to answer what if question imagine if you're a medical doctor if you know what happens if you give certain treatment how to choose the treatment will be very easy and if you are a sales manager and if you know what happens to the revenue if you give certain promotions the decision will also be very easy so in our project we would like to automate this process and optimize decisions with machine learning and in particular causal machine learning to answer such questions we need to understand choosing one is understand the relationship between different variables for example if we give like a membership out will this increase like the number of services on our Azure cloud and will this drive the revenue afterwards right so we want to understand the causal relationship and another thing we need to understand to final answer question is the consequence of the action how much is a revenue growth if you give a discount or if you give a membership and this would commonly call causal inference so both parts are needed so our team is actually the first team in the world that we Bridge causal Discovery causal inference and the Deep learning into a single framework so that will allow users to provide us historic data and we can answer all these questions tell them the relationship between variables tell them the consequences of different actions and be able to recommend optimize the decisions so there is a lot of research ongoing and this is a framework bringing us close to real world impact because it's a flexible scalable and be able to deal with all type of real world data and so in our mind in research this is at the dawn of a golden era so apart from Advanced research our team is really rooted in real world impact so the example we gave about sales we're actually working with our Microsoft Global Channel sales team to help them to optimize the engagement decisions to drive the revenue growth apart from the sales application we are in active conversation and collaboration Exploration with AF industry gaming Ai and many other departments in Microsoft and so of course with our team's aim is actually to empower everyone and every organization in the world to be able to do decision organization with causal machine learning so we're also exploring collaboration with platform level service for example cognitive service to make an even bigger impact so if you find that this type of research can help you or if you want to collaborate with us please contact us