This is really up to you and your imagination. We are hoping to see all sorts of imaginative ideas that mix the abilities of machine learning and messenger. We are intentionally trying not to give you too much guidance as we don’t want to influence your creativity! However, just to make it a little bit easier, here is one of the ideas that one of the teams will be working on: 
Problem: I walk past a historical statue or a building and I want to know more about it.
Solution: The ability to take a picture of the object using messenger and analyse it using machine learning so we can surface useful information about it to the user 

BACKGROUND

In 1966, Joseph Weizenbaum created a natural language processing computer program, named ELIZA, to demonstrate the superficiality of communication between humans and machines.
Ever since then, researchers and developers have been working on creating truly intelligent "chatbots" that understand human emotion and context.
Competitions like the Loebner Prize have been held, where chatbots undergo the Turing Test (https://en.wikipedia.org/wiki/Turing_test) to test their capabilities.
If the evaluator cannot reliably tell the machine from the human, the machine is said to have passed the test. The test results do not depend on the machine's ability to give correct answers to questions, only how closely its answers resemble those a human would give.
The techniques that power bots range from simple pattern matching, parsing, intent recognition, to more structured and sophisticated approaches like AIML (Artificial Intelligence Markup Language) and Markov chains.

References:
PyTorch Chatbot Tutorial: https://pytorch.org/tutorials/beginner/chatbot_tutorial.html
A Survey of Chabot Systemsthrough a Loebner Prize Competition: https://pdfs.semanticscholar.org/9447/1160f13e9771df3199b3684e085729110428.pdf

Prize: 

To be announced