Excellente synthèse des bouleversements induits par #ChatGPT par @nbouzou Nicolas Bouzou est l’un des TRÈS rares économistes qui a compris cette révolution
RESEARCH
-
Manfred Clynes Touch Profile and Haptic Interface Innovation
By
–
Reminds me of that last Manfred Clynes touch profile
-
John Schulman on Reinforcement Learning from Human Feedback
By
–
John Schulman – Reinforcement Learning from Human Feedback: Progress and Challenges https://
bit.ly/41NGSs0 #AI #MachineLearning #DeepLearning #LLMs #DataScience -
Groq ISC 2023 Workshop: DataFlow Architectures for GNN Workloads
By
–
Join @GroqInc for an ISC 2023 Workshop Paper presentation exploring the use of dataflow architectures for GNN workloads on May 25 at 2 pm. In collaboration with @Argonne
, we will demonstrate how DataFlow architectures such as GroqChip™ overcome inefficiencies of traditional GPUs -
Stanford Shows LLM Pre-training Cost Reduction Breakthrough
By
–
Excellent work from Stanford showing pre-training can be greatly cost-reduced. It is our collective hope that LLM training can become methodical, scalable, and efficient.
-
15-Year-Old AI Developer at Stanford Advancing Medical Research
By
–
Meet the 15-Year-Old AI App Developer Changing the World at Stanford Adrit Rao is not like most teenagers. At 15, he goes to high school. But he is also writing peer-reviewed medical papers for a Stanford internship, and more https://
bit.ly/3o7QvEe -
Self-Attention Mechanism in Large Language Models: Complete Guide
By
–
Understanding and Coding the Self-Attention Mechanism of Large Language Models From Scratch https://
bit.ly/3YL821l #AI #DeepLearning #MachineLearning #DataScience -
Cerebras Wafer-Scale Cluster Simplifies LLM Training
By
–
Large language models are extremely challenging to train. Our latest blog discusses how the Cerebras Wafer-Scale Cluster eliminates the need for huge compute budgets and complex distributed compute techniques. Read how we train models with a few clicks: https://
hubs.li/Q01R433-0 -
Hugging Face: Critical Infrastructure for AI-Powered Web Evolution
By
–
HF is increasingly a critical part of the AI model development infrastructure, and by extension (because everything will have these language and diffusion models in one way or another) what will go on to be the NEW modern web. (Since web3 is taken we can call it webGPT I guess.)
-
Steering GPT-2-XL With Activation Vector Techniques
By
–
Curious to learn more about the activation vectors @amasad is talking about at 25:00? https://
alignmentforum.org/posts/5spBue2z
2tw4JuDCx/steering-gpt-2-xl-by-adding-an-activation-vector
…