Discover AI Courses

Vibe Coding as a Learning Machine - Exploring Vibe Science and AI Research

1 review

Explore the concept of Vibe Coding as a learning machine and its potential application to scientific discovery, examining the CodeScientist preprint and discussing if AI can facilitate "Vibe Science."

• YouTube
• 37 minutes
• On-Demand
• Free Video

Mistral AI Magistral Medium - Causal Reasoning Test and Performance Evaluation

Discover Mistral AI's new Magistral models through live causal reasoning tests, comparing performance between open-source and enterprise versions with real-time coding challenges.

• YouTube
• 22 minutes
• On-Demand
• Free Video

WHY AI FAILS TO LEARN - Vision Language Models

Discover why vision-language models struggle with continual learning and explore Google DeepMind's innovative aligned model merging solution to overcome catastrophic forgetting.

• YouTube
• 25 minutes
• On-Demand
• Free Video

AI Thinking Illusion by Apple - Debunked

Explore AI reasoning capabilities through detailed performance testing of OpenAI o3, DeepSeek R1, and Claude Opus 4 to determine if AI can truly think or just mimics reasoning patterns.

• YouTube
• 36 minutes
• On-Demand
• Free Video

Layer Increasing Network Scaling (LiNeS) - Preventing Catastrophic Forgetting in Language Models

Explore LiNeS, a novel post-training layer scaling technique that prevents catastrophic forgetting in large language models while enhancing multi-task performance and model merging capabilities.

• YouTube
• 21 minutes
• On-Demand
• Free Video

Understanding Task Vectors in Vision-Language Models - Cross-Modal Representations

Explore how Vision-Language Models create task vectors - internal representations enabling cross-modal performance through text and images, revolutionizing AI's ability to understand and execute diverse tasks.

• YouTube
• 23 minutes
• On-Demand
• Free Video

Spectral Analysis of LoRA vs Full Fine-Tuning in Language Models

Delve into the technical comparison between LoRA and full fine-tuning methods for language models, exploring their structural differences, spectral properties, and impact on model performance.

• YouTube
• 28 minutes
• On-Demand
• Free Video

CODE RED: TTRL Unlocks AI Self-Evolution

Explore the boundaries of AI self-learning through TTRL methodology, examining the limits of self-rewarding and self-referencing reinforcement learning in language models.

• YouTube
• 33 minutes
• On-Demand
• Free Video

Off-Policy "Zero RL" Explained in Simple Terms

Discover how off-policy reinforcement learning compares to SFT for AI reasoning, exploring the LUFFY approach that integrates on-policy and off-policy zero RL for effective knowledge transfer without traditional RL methods.

• YouTube
• 46 minutes
• On-Demand
• Free Video

TEXAS: Fine-Tuning Is for Cowards - Do RL

Dive into the debate between Supervised Fine-Tuning and Reinforcement Learning for AI reasoning, exploring research findings on which approach yields better results for vision-language models.

• YouTube
• 33 minutes
• On-Demand
• Free Video

DeepSeek R1 0528 - 8B vs 671B Logic Reasoning Test

Explore the reasoning capabilities of DeepSeek R1 0528 models through a comparative logic puzzle test between the 8B distilled version and the full 671B model, revealing performance differences.

• YouTube
• 42 minutes
• On-Demand
• Free Video

RAG and Self-Evolving Agent ALITA with MCP Toolbox

Explore ALITA, a self-evolving AI agent that uses RAG-MCP for unscripted evolution, enabling scalable agentic reasoning with minimal predefinition and maximum adaptability.

• YouTube
• 20 minutes
• On-Demand
• Free Video

Optimizing LLM Action Planning Space with In-Context Learning

Discover how In-Context Learning (ICL) can optimize action planning for LLMs, enhancing AI systems' ability to determine effective action sequences for reaching objectives in virtual and real environments.

• YouTube
• 15 minutes
• On-Demand
• Free Video

Uncertainty AI: Breakthrough for Human-AI Reasoning

Discover how Google's Uncertainty AI framework in MedAI creates a blueprint for human-aligned AI by implementing iterative, transparent reasoning processes that mirror clinical decision-making in high-stakes domains.

• YouTube
• 14 minutes
• On-Demand
• Free Video

A Smarter Way to Fine-Tune LLMs - Unlock Deeper Reasoning

Discover a smarter approach to fine-tuning Large Language Models that enhances their reasoning capabilities and generalization from in-context learning, based on research from Google DeepMind.

• YouTube
• 19 minutes
• On-Demand
• Free Video