I got the lucky opportunity to go to Ohio State University HackAI event to judge and mentor students last weekend. The enthusiasm of the students towards AI was worth the trip alone. My teammates and I were able to guide students although they were self-driven with my team proving help only when needed. I wanted to share some of the highlights of the event and share some of the amazing work of the students.
- Student said it was his first time using in Azure and impressed with how easy it was, he said “it (Azure Machine Learning Studio) is freaking awesome!”
- The winner of the Microsoft challenge were high school students! They also mentioned that it was their first-time using Azure (and Azure Machine Learning) and they were amazed at how easy it was and really enjoyed using the toolset.
- Roughly 80 students at the start of the class, during the end of the day - there were 30 people (10 teams) that delivered full projects complete with presentations & demos within the 12-hour hackathon.
Day started out with breakfast and a keynote event where the students learned about how they were going to be evaluated and the expectations around judging. The Microsoft team went through introductions and presented our challenge on
- “Best AI hack with Azure Machine Learning Service and/or Azure ML Studio”
Goal: Use Microsoft Machine Learning technologies to deliver anything & everything to solve the students issues they come up with.
Grading criteria: AI functionality, Technicality, Originality, User Experience, and Teamwork
Students were excited to learn that Microsoft has put together easy Machine learning software described as “drag and drop machine learning” and that there were tutorials to get started.
Students quickly assembled to their stations where they were actively brainstorming ideas for their projects. We walked around just to introduce ourselves, learn about their ideas, and suggested tools within Microsoft that could be helpful to them.
Some of the initial project ideas we heard consisted of:
- Machine learning to find Waldo (Where’s Waldo)
- Using motion/camera to control gaming (1st person shooter)
- Question/Answer database for Ohio traffic laws using machine learning to retrieve the best answers
- Use Instagram with cognitive vision and ML to determine where parties are
- Note: remember we are talking about college students here :)
Initially there were more questions with our Microsoft team but then the students went into all out focus mode only stopping for pasta and pizza breaks.
Presentation / Judging
The list of remaining students went down to about 30 from 80 as the judging was about to begin. Some enjoyed the challenge but weren’t willing to present and some just had some other priorities like pending homework they had to get to 😊.
The 10 presentations were beyond fantastic, and we really were impressive for what a team can accomplish in 12 hours. Here is a list of project descriptions with respect to the awards that were delivered.
1st Place Overall
RacingGamingControl – Team put together a racing game and were able to demonstrate being driven by voice commands and camera-based motion gestures.
2nd Place Overall
MODOFE – Team created blurring effect for cheap cellphones. Estimate depth of an image using deep learning.
1st Place Impact Award
Predicting Molecular Energy - Determined molecular energy to engineer medicine/drugs. Measured vibrations using Azure Machine Learning to enable predictions.
2nd Place Impact Award
SocialAI – Platform to make recommendations for events/activities and extract hobbies based on your pictures and locations. Use deep learning model to suggest events. Example: Sporty pictures -> recommend cardio, boxing lessons.
1st Place Creativity Award
Genetic Battebots – Genetic algorithms (Theory of Evolution/Natural Selection) take battle and pair against each other using machine learning to determine outcomes.
2nd Place Creativity Award
Woosh – Tinder search using an uploaded user picture and comparing to a database of extracted pictures (4,000 profiles within 15 miles) using cognitive vision to determine if the user is on tinder or not.
Microsoft Challenge Award
Winner - Predicting Molecular Energy
Runner up – Woosh
Toxicity Meter – Detecting sentences if they are toxic or not (using bad language for example). Demonstrated with key logger as user typed and a windows popup for the alert detection.
Othello – Team modeled machine learning with Othello. Agent learned against itself to build model (Reinforcement learning). Trained 84,000 times against itself. Could play against a human in console-based demo.
Model Comparison – Team took breast cancer data to see if patients were benign or not using machine learning.
NLP Game engine – CNN model to implement game engine. Ideas focused on how game programming is hard but if you know English you could type natural commands to create elements in a game and motion.
Everyone did a great job presenting! My team and I were delighted to be a part of the event and look forward to seeing more student contributions to AI in the future! Thanks again to Microsoft for sponsoring the event and allowing us to travel and interact with students!
Hack AI is sponsored by OHI/O, a program dedicated to fostering tech culture at the Ohio State University.
Our Microsoft Team (pictured above): Mayank, Lilong, and Me (Nate)