Developmental Machine Learning: From Human Learning to Machines and Back
October 17-21, 2022
Recent advances in artificial intelligence, enabled by large-scale datasets and simulation environments, have resulted in breakthrough improvements in areas like object and speech recognition, 3D navigation, and machine translation. In spite of these advances, modern artificial learning systems still pale in comparison to the competencies of young human learners. The differences between human learning and the paradigms that currently guide machine learning are striking. For example, children actively identify both the concepts to be learned and the data items used for learning, they infer the labels for learning from ambiguous perceptual data, and they learn from continuous streams of percepts without storing and curating large datasets. Artificial intelligence researchers are increasingly looking to developmental science for ideas and inspiration to improve machine learning, while developmental scientists are adopting tools from data science and machine learning to analyze large datasets and gain insights into developmental processes. This Dagstuhl Seminar will provide an opportunity to catalyze new connections between the developmental and machine learning research communities by bringing together researchers in linguistics, psychology, cognitive science and neuroscience with investigators working in computer vision, machine learning and robotics (full text).
Jim Rehg
Pierre-Yves Oudeyer
Linda Smith
Sho Tsuji
Schedule
Note to speakers: Please limit your talk to 15 minutes.
Each talk is followed by a moderated 5 minute Q&A.
Day 1: Monday, October 17
Morning Session
9:00 - 9:20 - Intro + Goals
9:20 - 10:05 - Tutorial on Developmental Learning - Michael Frank - Word learning as a case study for children’s learning abilities
10:05 - 10:30 - Discussion
10:30 - 11:00 - Coffee Break
11:00 - 11:45 - Tutorial on Machine Learning - Andrew Zisserman + Jim Rehg
11:45 - 12:15 - Discussion
12:15 - 1:00 - Lunch
Afternoon Session
1:00 - 1:20 - Felix Hill - How language can help machines to acquire general intelligence
1:20 - 1:40 - Alejandrina Cristia - Using machine learning in early language acquisition research: Examples from long-form audio-recordings
1:40 - 2:00 - Michael Frank - Predictive models of early language learning
2:00 - 2:20 - Pierre-Yves Oudeyer - Language and Culture Internalization for Autotelic Human-Like AI
2:20 - 2:40 - Emmanuel Dupoux - Simulating early language acquisition using self-supervised learning
2:40 - 3:00 - Discussion
3:00 - 3:30 - Tea break
3:30 - 3:50 - Daniel Swingley - Rethinking the developmental pathway of early infant language learning
3:50 - 4:10 - Rhodri Cusack - Can Machine Learning Inform the Science of Infant Development, and Vice-Versa?
4:10 - 4:30 - Abdellah Fourtassi - ML as a tool” vs. “ML as a model” for the study of child development in the wild
4:30 - 4:50 - Jitendra Malik (Remote) - Learning Vision for Walking
4:50 - 5:10 - Discussion
5:10 - 6:00 - Break before dinner
6:00 - Dinner
Day 2: Tuesday, October 18
Morning Session
9:00 - 10:30 - Working Group Time
10:30 - 11:00 - Coffee Break
11:00 - 12:15 - Working Group Time
12:15 - 1:00 - Lunch
Afternoon Session
1:00 - 1:20 - Andrew Zisserman (Remote) - Audio-visual self-supervised learning
1:20 - 1:40 - Linda Smith (Remote) - Why self-generated behavior has more radical consequences that you might originally think
1:40 - 2:00 - Jim Rehg - Connecting 3D Shape Learning and Object Categorization
2:00 - 2:20 - Judy Hoffman - The Impact of Dataset Bias on Model Learning
2:20 - 2:40 - Olivier Sigaud - Towards Teachable Autonomous Agents: How can developmental psychology help?
2:40 - 3:00 - Discussion
3:00 - 3:30 - Tea break
3:30 - 3:50 - Casey Lew-Williams - The First 1,000 Days Project
3:50 - 4:10 - Uri Hasson - The First 1,000 Days Project
4:10 - 4:30 - Maithilee Kunda - Studying infant-like visual category generalization using the Toybox dataset
4:30 - 4:50 - Marc'Aurelio Ranzato - The Never-Ending VIsual classification Stream (NEVIS) 1.0
4:50 - 5:10 - Atsushi Nakazawa - Does Affective communication increase the relation between children with ASD and their mothers?
5:10 - 5:30 - Discussion
5:30 - 6:00 - Break before dinner
6:00 - Dinner
Day 3: Wednesday, October 19
Morning Session
9:00 - 10:00 - Working Group Time
10:00 - 10:30 - Coffee Break
10:30 - 11:50 - Working Group Time
11:50 - 12:15 - Discussion
12:15 - 1:00 - Lunch
Afternoon Session
1:00 - 5:00 - Social Outing
6:00 - Dinner
Day 4: Thursday, October 20
Morning Session
9:00 - 9:20 - Sho Tsuji (Remote) - SCALa: A blueprint for computational models of language acquisition in social context
9:20 - 9:40 - Eon-Suk Ko - Enhancement of cues and the oddball effect in child-directed speech
9:40 - 10:00 - Anne Warlaumont - Temporal patterns in vocal even sequences produced by human infants and computational vocal learning models
10:00 - 10:20 - Discussion
10:20 - 11:00 - Coffee Break
11:00 - 11:20 - Rebecca Saxe - Human infants' brains are specialized for social functions
11:20 - 11:40 - Chen Yu - Magnifying Time and Space: New Ways of Studying Early Development and Learning from the Infant’s Point of View
11:40 - 12:00 - Hana D'Souza - Towards embracing complexity to understand atypical development: the case of Down syndrome
12:00 - 12:20 - Discussion
12:20 - 1:00 - Lunch
Afternoon Session
1:00 - 1:20 - Celeste Kidd - Truth, lies, and misinformation during cognitive development
1:20 - 1:40 - Gert Westermann - Curiosity in infants and computational models
1:40 - 2:00 - 4 talks - Post/Pre Doctoral Session 1 - Stefan Stojanov, Anh Thai, Hiromichi Hagihara, Clement Romac
2:00 - 2:10 - Short Q&A
2:10 - 2:30 - 4 talks - Post/Pre Doctoral Session 2 - Jelena Sucevic, Maureen de Seyssel, Thomas Carta, Marvin Lavechin
2:30 - 2:40 - Short Q&A
2:40 - 3:00 - Discussion
3:00 - 3:30 - Tea break
3:30 - 3:50 - Ingmar Visser - Visual attention development in infancy
3:50 - 4:10 - David Crandall - Studying visual object learning with egocentric computer vision
4:10-4:30 - Kristen Grauman - Visual affordances from video: learning to interact by watching people
4:30 - 4:50 - Discussion
4:50 - 6:00 - Working Group Time
6:00 - Dinner
Day 5: Friday, October 21
Morning Session
9:00 - 10:30 - Report back on conclusions from each working group
10:30 - 11:00 - Coffee Break
11:00 - 12:15 - Discussion about the next steps for this community
12:15 - 1:00 - Lunch
1:00 - Departure