Philia is the name of our emotive chatbot and is our fundamental research project about injecting emotions into AI. We started this fundamental research project with beginners. Yes, you read that right: fundamental research done by (almost) no experts. But guess what: we managed to make it work! It was not easy, it was not efficient but we had a lot of fun! And we can definitively say that the two first cohorts from ConcordAI were very successful. Thank you all!

The story so far

  • 2021-01-01: We are completely rewriting the overall architecture.
  • 2020-05-01: Academic research started at École Polytechnique with Pr. Michel Gendreau.
  • 2020-01-01: Working with a small team in stealth mode...
  • 2019-08-01: MVP shown at ConcordAI Demo Night. Despite our MVP containing bugs, not totally working as expected and insufficiently trained, people enjoyed the demo of a chatbot with some personality.
  • 2019-05-05: Second cohort from ConcordAI working on the project (May 2019 - July 2019). Many had heard about our first success. This cohort of ten people had some knowledge of ML. Interestingly and despite our clear intentions from the very beginning, we had to demonstrate that Deep Learning was not enough to do what we wanted to do. Three people left after the first day. Luckily, the rest of the group stayed and enjoyed the ride. This time around, we went a little further into the details and constructed a fully functional emotional chatbot: our first MVP (Minimum Viable Product). Along the way, we realized that we needed to redefine some standards in linguistic.
  • 2019-03-30: POC shown at ConcordAI Finale Far Into the Future: Implications of a World Run by AI. Even at this early stage, our POC dispalyed a behavior that is impossible to obtain using only ML. Our system showed some emotions but it also could decide to avoid certain reactions that were not aligned with a given emotion.
  • 2019-01-12: First cohort from ConcordAI working on the project (January 2019 - March 2019). Only 7 persons without much knowledge of ML or OR joined us, probably because those in the know didn't believe we could achieve what we claimed. We took five full sessions just to explain the main ideas but the remaining 7 sessions were fruitful and we managed to construct a POC (Proof of Concept)! In the middle of the session, we were joined by a CHO (Chief Happiness Officer). Everybody stayed on board until the end and the motivation ran high!