Browse Lessons

The Model Editing Problem #machinelearning #ai #podcast

Localizing and Editing Knowledge in LLMs with Peter Hase - 679

Will it ever be possible to stop jailbreaks? #machinelearning #ai #podcast

Are LLMs too vulnerable to be used as agents? #machinelearning #ai #podcast

Coercing LLMs to Do and Reveal (Almost) Anything with Jonas Geiping - 678

Video is better than language at controlling AI robots #machinelearning #ai #podcast

Open-source AI makes deepfakes more accessible. #machinelearning #ai #podcast

Is OpenAI changing its model behind the scenes? #machinelearning #ai #podcast

Akshita finds an issue in PyTorch #machinelearning #ai #podcast

Emergent behaviors in LLMs depend on your evaluation metric #machinelearning #ai #podcast

How risky are open-source LLMs really? #machinelearning #ai #podcast

Joint Embedding Architectures simplify AI reasoning #machinelearning #ai #podcast

V-JEPA might be the future of AI reasoning #machinelearning #ai #podcast

V-JEPA, AI Reasoning from a Non-Generative Architecture with Mido Assran - 677

Video generators are world models #machinelearning #ai #podcast

Video as a Universal Interface for AI Reasoning with Sherry Yang - 676

Assessing the Risks of Open AI Models with Sayash Kapoor - 675

What's in your Big Data? #artificialintelligence #ai #opensource

OLMo: Everything You Need to Train an Open Source LLM with Akshita Bhagia - 674

Teaching AI to think step by step #artificialintelligence #ai #reasoning

Training Data Locality and Chain-of-Thought Reasoning in LLMs with Ben Prystawski - 673

Fine Tuning is key to building practical AI tools

Reasoning Over Complex Documents with DocLLM with Armineh Nourbakhsh - 672

Are Emergent Behaviors in LLMs an Illusion? with Sanmi Koyejo - 671

Are GPUs enough to run AGI? #ai #podcast #machinelearning #reinforcementlearning

Reinforcement learning is replacing control algorithms. #ai #reinforcementlearning

Generative AI is making RL agents practical

The challenge of keeping vector databases up to date

Vector databases are the evolution of search engines

Vector databases allow startups to compete with SOTA

RAG makes LLMs better, even on data they have already been trained on. #ai #ML #Pinecone #RAG

AI Trends 2024: Reinforcement Learning in the Age of LLMs with Kamyar Azizzadenesheli - 670

Building and Deploying Real-World RAG Applications with Ram Sriharsha - 669

Rapid AI advancements amplify social gaps like underrepresentation and bias, magnifying disparities

Nightshade: a poison pill that subtly distorts image composition, causing model confusion & collapse

AI-generated art imitates artists’ style and artworks, robbing them of their identity and livelihood

Nightshade: Data Poisoning to Fight Generative AI with Ben Zhao - 668

The main challenge comes from difficulties in discrete optimization. Scaling needs improved strategy

Writing the expressive transformer program solves tasks despite discrete optimization limitations

Mechanistic interpretability's goal is to reverse engineer the model to human-readable algorithms

Learning Transformer Programs with Dan Friedman - 667

The emergence of capabilities is influenced by training data & patterns discovered as data scales up

Better tools & open-source models enable the academic community to understand model's complex system

Build robust system for an open world that can manage novelty & classify LLMs to avoid hallucination

AI Trends 2024: Machine Learning & Deep Learning with Thomas Dietterich - 666

Tune in to know what are the most exciting opportunities to look out for in computer vision!

Vision community faces evaluation challenges and should lean on cost-effective automatic evaluation

LangChain vs LlamaIndex. Beyond data, constructing context & post-filtering enhance accuracy from VB