True Positive Weekly #70

The most important artificial intelligence and machine learning news and articles

Hey, in this issue: inference of Llama 2 in one file of pure Python; a simple way to regulate powerful AI models; instruction-tune Llama 2; a generative AI extension for JupyterLab; eliminating hallucinations in large language models; coding the self-attention mechanism from scratch; and more.

• [Project] Llama2.py: inference of Llama 2 in one file of pure Python

• NVIDIA releases ‘Neuralangelo’ open source code

• [Opinion] Here’s a simple way to regulate powerful AI models

• [NYT] AI can’t build a high-rise, but it can speed up the job

• [Tutorial] Instruction-tune Llama 2

• [Tool] Jupyter-ai: a generative AI extension for JupyterLab

• [Tutorial] Eliminating hallucinations in large language models with finite state machines

• [Tutorial] Coding the self-attention mechanism of large language models from scratch

Enjoy the newsletter? Help us make it bigger and better by sharing it with colleagues and friends.