Video essays on AI research and interviews with leading scientists in the field.
Linking important ideas of AI. The border between high-level and technical thinking.

Self-play for Self-driving and where Scaling Reinforcement Learning is Heading with Eugene Vinitsky
1:09:23
Self-play for Self-driving and where Scaling Reinforcement Learning is Heading with Eugene Vinitsky
An Unexpected Reinforcement Learning Renaissance
39:49
An Unexpected Reinforcement Learning Renaissance
OLMo leads on the secrets of training language models (w Dirk Groeneveld, Kyle Lo, & Luca Soldaini)
1:12:43
OLMo leads on the secrets of training language models (w Dirk Groeneveld, Kyle Lo, & Luca Soldaini)
How language model post-training is done today
53:51
How language model post-training is done today
Quick recap on the state of language model reasoning
16:22
Quick recap on the state of language model reasoning
Finbarr Timbers on the Future of Reinforcement Learning
1:08:33
Finbarr Timbers on the Future of Reinforcement Learning
Tim Dettmers on Open-source AI, LMs, SWE Bench, Agents, Quantization, & Optimization
1:15:46
Tim Dettmers on Open-source AI, LMs, SWE Bench, Agents, Quantization, & Optimization
Andrew Carr on Pushing the Boundaries of Generative AI (Beyond Text)
54:10
Andrew Carr on Pushing the Boundaries of Generative AI (Beyond Text)
Arvind Narayanan on making sense of AI hype
54:22
Arvind Narayanan on making sense of AI hype
Andrew Trask on changing how LLMs store and access information
1:00:12
Andrew Trask on changing how LLMs store and access information
Riley Goodside on OpenAI's o1 and the frontier of prompting LLMs
1:08:39
Riley Goodside on OpenAI's o1 and the frontier of prompting LLMs
Llama 3.2 vision vs. Molmo: Open multimodal models
14:04
Llama 3.2 vision vs. Molmo: Open multimodal models
OpenAI's Strawberry and Spending More on Inference
10:40
OpenAI's Strawberry and Spending More on Inference
OLMoE and training better foundation models
10:31
OLMoE and training better foundation models
Defining open-source AI and open data
8:01
Defining open-source AI and open data
Interviewing Ross Taylor on LLM reasoning, Llama post training, and Galactica
1:02:22
Interviewing Ross Taylor on LLM reasoning, Llama post training, and Galactica
Interviewing Sebastian Raschka on the state of open LLMs, Llama 3.1, and AI education
1:03:42
Interviewing Sebastian Raschka on the state of open LLMs, Llama 3.1, and AI education
Llama 3.1 405b and Meta's open source AI strategy
15:22
Llama 3.1 405b and Meta's open source AI strategy
Switching to Claude 3.5 from ChatGPT
6:40
Switching to Claude 3.5 from ChatGPT
Interviewing Dean Ball on AI policy: SB-1047, Llama 405B, scaling laws, China, and more
56:31
Interviewing Dean Ball on AI policy: SB-1047, Llama 405B, scaling laws, China, and more
Frontiers in LLM synthetic data
11:27
Frontiers in LLM synthetic data
How Apple Intelligence Works
12:35
How Apple Intelligence Works
A realistic path to robotic foundation models
7:49
A realistic path to robotic foundation models
There is no LLM data wall
8:29
There is no LLM data wall
OpenAI vs. Scarlett Johansson: Name, image, and AI's likeness
9:03
OpenAI vs. Scarlett Johansson: Name, image, and AI's likeness
OpenAI builds Her
12:28
OpenAI builds Her
OpenAI's RLHF Specifications
14:05
OpenAI's RLHF Specifications
RLHF: A thin line between useful and lobotomized
13:08
RLHF: A thin line between useful and lobotomized
Phi 3 and Arctic: Outlier LMs are hints
9:46
Phi 3 and Arctic: Outlier LMs are hints
AGI is what you want it to be
10:38
AGI is what you want it to be