Videos
See AI2's full collection of videos on our YouTube channel.Viewing 31-40 of 216 videos
Synthetic Data for Language-Guided Agents
March 18, 2022 | Peter AndersonRecent advances in vision and language modeling have been powered by truly massive datasets, often mined from the web. Instruction-following robots will also require large amounts of data to train and evaluate. However, images/videos/documents found on the web do not satisfy the needs of embodied agents, and data…
Soft Robotics and AI towards Embodied Intelligence
March 4, 2022 | Fumiya IidaSoft robotics research has made considerable progress in many areas of robotics technologies based on deformable functional materials, including locomotion, manipulation, and other morphological adaptation such as self-healing, self-morph, and mechanical growth. While these technologies open up many new robotics…A Flexible Framework for Machine Learning
March 2, 2022 | Ferran AletIn this last decade, we have seen a lot of progress in AI and Machine Learning using different variations on a single recipe: we specify a task as learning a function mapping inputs to outputs and we train a single neural network to approximate it. In this talk, I will show that this one NN per task framework can…
The Future is Hear: Advances in Computational Audition and Sound Manipulation
December 1, 2021 | Bryan PardoAbstract: Computer Audition is the sonic analog to Computer Vision and encompasses much more than just speech to text. Northwestern University’s Interactive Audio Lab (IAL), headed by Prof. Bryan Pardo, is a world leader in Computer Audition. IAL develops new techniques and technologies for identifying sound…
Explaining Answers with Entailment Trees
November 12, 2021 | Bhavana DalviOur goal, in the context of open-domain textual question-answering (QA), is to explain answers by showing the line of reasoning from what is known to the answer, rather than simply showing a fragment of textual evidence (a "rationale'"). If this could be done, new opportunities for understanding and debugging the…
Adapting to Long Tail Domains: A Case Study in Clinical Information
November 4, 2021 | Aakanksha NaikAdvances in deep learning, especially self-supervised representation learning, have produced models that reach human parity on many benchmark datasets, which cover a variety of natural language understanding tasks. However, benchmark datasets are constructed from naturally occurring texts, and are no exception to…
Beyond the Label: Robots that Reason about Object Semantics | Embodied AI Lecture Series
October 15, 2021 | Sonia ChernovaReliable operation in everyday human environments – homes, offices, and businesses – remains elusive for today’s robotic systems. A key challenge is diversity, as no two homes or businesses are exactly alike. However, despite the innumerable unique aspects of any home, there are many commonalities as well…
GooAQ: Open Question Answering with Diverse Answer Types
October 12, 2021 | Daniel KhashabiBased on the following paper: https://arxiv.org/abs/2104.08727
Open-Ended Learning Leads to Generally Capable Agents | Embodied AI Lecture Series
October 1, 2021 | Max JaderbergIn this talk I will cover our recent publication "Open-Ended Learning Leads to Generally Capable Agents" (https://deepmind.com/blog/article/generally-capable-agents-emerge-from-open-ended-play). In this work we turn our attention to how to create embodied agents in simulation that can generalise to unseen test…
Computing with a Mess | Embodied AI Lecture Series
September 17, 2021 | Stefan MihalasComputing with a mess: how nonstationary, heterogeneous and noisy components help the brain’s computational power While artificial neural networks have taken inspiration from biological ones, one salient difference exists at the level of components. Biological neurons and synapses have heterogeneous transfer…