cpapadimitriou
commented
7 years ago Thanks for the feedback, Seung. I went back and changed the spots where I was talking about soulmate/ultimate partner. You are right, Love-with-AI is used to suggest successful dating partners closer to the user's type, rather than finding the user's soulmate.
To your second point, these are some very interesting questions that you are raising. My idea was to perform the deep learning algorithm on the fly based on data collected for each individual (and maybe also use similarity algorithms to learn from individuals with similar "taste"), but of course it is a little tough for me to suggest exactly how it would work since I have not implemented these technologies yet. I am super excited to learn more about deep learning in W266 and maybe then I will have better answers to your questions :)
@cpapadimitriou
I think your idea is using machine learning with image and historical conversation data to suggest a better dating partner in a dating app.
What should be kept: I think you should definitely keep the first sentence in the middle. This clearly defines how you want to solve the problem in research design standpoint. Personally, I like this line: "But the truth is that people are not always attracted to people that are similar to them".
What could be cut: To me, somehow, "type" and "soulmates" sound like two different things. Love-with-AI sounds more like finding your type instead of your soulmate.
What could be rearranged: Your idea flows well.
What could be added: I'm curious how you will make machines learn. Once a user clicks "Like", does the app perform deep learning algorithm on fly? Would it be fast enough to do that? Also, for text analysis, how would you define "interesting conversations"? It sounds like Love-with-AI will have a model for each individual, but do you think we can collect enough data from each individual to build customized machine learning models?
This is a very interesting topic, and it will definitely attract a lot of people.