Nicolas Gontier
Teaching Machines to Learn
Updates:
[05/2023] Sucessfully defended my PhD Thesis with the jury recommendation for the best thesis award at Polytechnique Montreal.
[10/2022] I am thrilled to join ServiceNow Research as a full-time Research Scientist in the Human-Machine Interaction Trough Language Lab.
About:
I am a Research Scientist at ServiceNow research. My research lies at the intersection of Natural Language Processing (NLP) and Reasoning (I like to call it "Natural Language Reasoning"). I am especially interested in Large Language Models (LLMs) that use external tools and language generation systems such as conversational agents.
I obtained a Ph.D. with the jury recommendation for the best thesis award at Polytechnique Montreal in May 2023 while being affiliated with the Mila (Montreal AI institute). Supervised by professor Christopher Pal, my research focused on proof generation and multi-step reasoning with Transformer Language Models. Prior to that, I obtained a Master of Science (M.Sc.) in Computer Science in association with the Reasoning and Learning Lab at McGill University, in 2018. Being supervised by professor Joelle Pineau, my research focused on dialogue systems. Before that, I obtained a Bachelor of Science (B.Sc.) in Computer Science from McGill University in 2016. I accompanied my Computer Science major with a minor in Economics.
I am also curious about a broad range of other topics such as AI Ethics, Finance, theoretical physics, AI in medicine, neuroscience, computational biology, software development, mobile apps development, computer security, and anything that is related to multimedia (3D animations, movie making, music).
Publications:
"Does Entity Abstraction Help Generative Transformers Reason?" ; Nicolas Gontier, Siva Reddy, Christopher Pal ; In Transactions of Machine Learning Research (TMLR) ; 2022 ; [paper] [openreview]
"Measuring Systematic Generalization in Neural Proof Generation with Transformers"; Nicolas Gontier, Koustuv Sinha, Siva Reddy, Christopher Pal; In Advances in Neural Information Processing Systems 33 (NeurIPS 2020) ; 2020 ; [paper] [slides] [poster] [code]
"Ethical Challenges in Data-Driven Dialogue Systems"; Peter Henderson, Koustuv Sinha, Nicolas Gontier, Nan Rosemary Ke, Genevieve Fried, Ryan Lowe, Joelle Pineau ; In proceedings of the AAAI conference on Artificial Intelligence, Ethics, and Society (AIES) ; 2017 ; [paper] [slides] [poster] [website] [code]
"Towards an Automatic Turing Test: Learning to Evaluate Dialogue Responses"; Ryan Lowe, Michael Noseworthy, Iulian Serban, Nicolas Gontier, Yoshua Bengio, Joelle Pineau ; In proceedings of the 55th annual meeting on Association for Computational Linguistics (ACL) (Outstanding Paper Award); 2017 ; [paper]
Master's Thesis:
"The Reasoning and Learning Lab Chatbot: a solution to the conversational intelligence challenge"; Nicolas Gontier, Joelle Pineau; McGill Master's thesis 08/2018; [pdf]
Project Description: Ensemble model with message scoring and selection mechanisms submitted at the NeurIPS 2017 Conversational Intelligence Challenge; 06/2017 - 09/2018; [paper] [code] [poster] [slides]
Projects:
The Rhyming Chatbot; Nicolas Gontier, Chris Pal, David Usher; Trained hierarchical RNNs and GPT models with systematic rhyming; Deployed as a Slack bot at ElementAI; 2019
On the Sensitivity of RNN decoders; Nicolas Gontier, Chris Pal; Evaluated the impact of different training techniques on RNN decoder sensitivity; 05/2019; [paper] [poster]
The RLLChatbot: a solution to the ConvAI challenge; Nicolas Gontier, Koustuv Sinha, Peter Henderson, Iulian Serban, Prasanna Parthasarathi, Michael Noseworthy, Joelle Pineau; Ensemble model with message scoring and selection mechanisms submitted at the NeurIPS 2017 Conversational Intelligence Challenge; 06/2017 - 09/2018; [paper] [code] [poster] [slides]
Caption to Image: Nicolas Gontier, Joshua Romoff, Prasanna Parthasarathi; Variational Encoder-Decoder trained to read math operations and produce the result as an MNIST image; 11-12/2017; [report] [slides] [code]
Ubuntu Dialogue Corpus with HRED and attention: Nicolas Gontier, Joumana Ghosn; Investigating the generic response generation problem with HRED models at Nuance Communications Summer Internship; 05-09/2017; [report] [poster]
Policy Gradient Methods for Dialogue Response Generation: Michael Noseworthy, Nicolas Gontier; Investigation of the ADEM score as a Dialog evaluation metric with the REINFORCE algorithm; 2017; [report] [slides]
Training a Discriminator to Compare Generative Dialogue Models: Nicolas Gontier; Trained RNNs to classify different dialogue responses (random, tf-idf, HRED, VHRED, true); 2017; [report] [slides] [code]
Sentiment Analysis on Movie Reviews: Nicolas Gontier; Random Forest and LSTM on Movie Reviews; 2016; [report] [slides] [code]
Advisor:
Scientist in Residence at Creative Destruction Lab - Montreal: Gave AI track interviews for the Fall 2021, Fall 2022 cohort.
AI Advisor at TheOneAI: Hired & suppervised AI engineers for the development of the Visual Search feature ("Visual Match").
Teacher Assistant:
Reviewer
ICLR: 2019, 2020, 2021, 2023
NeurIPS: 2021, 2022, 2023
TMLR: 2023