YouTube Courses

Intuitive Reasoning as Unsupervised Neural Generation

Explore intuitive reasoning in AI through neural generation, covering topics like common sense, language models, and multimodal information with researcher Yejin Choi.

• YouTube
• 1 hour 8 minutes
• Self-Paced
• Free Video

Jacob Andreas - Natural Language Explanations of Deep Networks

Explore strategies for generating natural language explanations of deep networks, covering medical diagnosis, multiagent communication, and neuron descriptions in AI systems.

• YouTube
• 1 hour 13 minutes
• Self-Paced
• Free Video

The Lottery Ticket Hypothesis - Michael Carbin

Explore the Lottery Ticket Hypothesis for efficient neural networks. Learn about pruning techniques, sparse architectures, and potential improvements in machine learning methods.

• YouTube
• 1 hour 6 minutes
• Self-Paced
• Free Video

Diverse Data and Efficient Algorithms for Robot Learning

Explore diverse data collection and efficient algorithms for robot learning, focusing on large-scale approaches and reinforcement learning for deformable object manipulation.

• YouTube
• 56 minutes
• Self-Paced
• Free Video

MIT EI Seminar - Phillip Isola - Emergent Intelligence- Getting More Out of Agents Than You Bake In

Explore emergent intelligence in AI through unsupervised learning, generative modeling, and self-assembling agents. Discover surprising outcomes beyond explicit programming in representation, video generation, and creature development.

• YouTube
• 58 minutes
• Self-Paced
• Free Video

Aligning Language Models with LESS Data and Simple Preference Optimization (SimPO)

Discover innovative approaches to language model alignment through LESS data selection and SimPO optimization, achieving superior performance with reduced computational resources and simplified objectives.

• YouTube
• 1 hour
• Self-Paced
• Free Video

Building General Zero-Shot Robot Policies - How Large Models Changed Robotics

Explore how large AI models revolutionize robotics through Task and Motion Planning, examining PaLM-E, RT-2, and Aloha Unleashed for advanced robot decision-making and dexterity.

• YouTube
• 1 hour 3 minutes
• Self-Paced
• Free Video

Redefining Context for Powerful Test-Time Adaptation Using Unlabeled Data

Explore advanced test-time adaptation techniques for foundation models, focusing on context-driven approaches to improve model performance across different domains using unlabeled data and inductive biases.

• YouTube
• 48 minutes
• Self-Paced
• Free Video

On Building General, Zero-Shot Robot Policies - EI Seminar

Explore the development of zero-shot robot policies for real-world tasks. Gain insights into Robot Utility Models (RUMs) and their ability to generalize to new environments without finetuning, achieving 90% success in novel settings.

• YouTube
• 54 minutes
• Self-Paced
• Free Video

Recent Progress on Foundation Model Supervision for Robot Learning

Discover how foundation models from non-robot domains can revolutionize robot learning, enabling automated task acquisition without extensive data collection or manual controller design.

• YouTube
• 53 minutes
• Self-Paced
• Free Video

The Promises and Pitfalls of Open-source Agent Systems

Explore the potential and challenges of open-source AI agent systems, examining their performance compared to closed-source alternatives and key issues in evaluation and generalization.

• YouTube
• 58 minutes
• Self-Paced
• Free Video

Games and Filters: A Road to Safe Intelligence

Explore game-theoretic approaches and safety principles in AI systems, from robotic control to human-AI interactions, focusing on developing reliable safety guarantees for autonomous systems.

• YouTube
• 49 minutes
• Self-Paced
• Free Video

Towards AI-Assisted Video Editing - Generating Shorts from Long Videos

Explore AI-powered video editing techniques that automatically transform long videos into engaging short content using machine learning models and multimodal LLMs.

• YouTube
• 1 hour 6 minutes
• Self-Paced
• Free Video

Smaller, Stronger, and Duration-Scalable Audio Learners

Explore state-space models for audio processing with DASS, a model that outperforms Transformers on AudioSet while being smaller and capable of handling hour-long recordings, unlike traditional models limited to short clips.

• YouTube
• 31 minutes
• Self-Paced
• Free Video

Giving Sight to Speech Models

Discover how Whisper-Flamingo integrates visual lip features into speech recognition models, improving performance in noisy conditions for both English recognition and multilingual translation.

• YouTube
• 24 minutes
• Self-Paced
• Free Video