Link to Youtube: https://youtu.be/w8S8sSPHqLw

Formula 1 recap (0:00)

State of AI: a year since gpt3 (9:30)

Origins of Deep learning and Generative AI (12:30)

Intelligence is compute power and model size (19:00)

Gemini, and how to define AGI (29:00)

How ChatGPT thinks (34:30)

Multimodality (40:20)

AI based robotics (46:30)

Impacts on society: Separating Alignment and Safety (49:40)

What we should train deep learning models for (55:50)

Links

Shoutout to rockerpoweredmohawk (F1 related content): https://www.youtube.com/@Rocketpoweredmohawk

The Gemini Lie: https://www.youtube.com/watch?v=90CYYfl9ntM&pp=ygUbZmlyZXNoaXAgY29kZSByZXBvcnQgZ2ltaW5p

Power of Multimodal AI (hints towards AGI)

https://youtu.be/ex1GeX0IhJQ?si=hrYWTE7BREbCszUn

Hinton's whitepaper (backpropagation applied on multi-layer perceptron in 1986): https://www.nature.com/articles/323533a0

LeNet whitepaper (mathematical proof of CNNs in 1990): https://direct.mit.edu/neco/article-abstract/1/4/541/5515/Backpropagation-Applied-to-Handwritten-Zip-Code?redirectedFrom=fulltext

AlexNet whitepaper (first popular practical application of CNNs in 2011): https://dl.acm.org/doi/10.1145/3065386

The paper that "transformed" the natural language processing world: https://dl.acm.org/doi/10.5555/3295222.3295349

OpenAI being inconsistent with model architecture transparency: https://en.wikipedia.org/wiki/GPT-4#Criticisms_of_transparency

Andrej Karpathy OpenAI pipeline talk - https://youtu.be/bZQun8Y4L2A?si=Im-cCSfGXhG87Yvz

Alpha fold (Predicting protein structure from sequence) - https://alphafold.ebi.ac.uk/

GNoME (Material Science AI) - https://deepmind.google/discover/blog/millions-of-new-materials-discovered-with-deep-learning/