Weeklog for Week 51: December 18 to December 24

Dec 24, 2023

Progress

Last work week this year, so finishing up some stuff.

I held a second AI workshop, this time a bit more technical. We only had two hours, so I couldn't cover as much as I wanted to. Talk to me about it!

Articles

The Most Borgesian Story - by Salty Spittoon
[2304.13712] Harnessing the Power of LLMs in Practice: A Survey on ChatGPT and Beyond: Beautiful lineage tree for GPT models
Bruce Sterling @bruces: "#AI " - Mastodon
Autonomous Subs Use AI to Wayfind Without GPS - IEEE Spectrum
ArtBot - Create images with Stable Diffusion, utilizing the AI Horde
Church vs Curry Types
Interactive code cells
The Cheap Web
Ventoy: "Ventoy is an open source tool to create bootable USB drive for ISO/WIM/IMG/VHD(x)/EFI files. With ventoy, you don't need to format the disk over and over, you just need to copy the ISO/WIM/IMG/VHD(x)/EFI files to the USB drive and boot them directly."
Introducing Fly Kubernetes · The Fly Blog
pchunduri6/rag-demystified: An LLM-powered advanced RAG pipeline built from scratch
unslothai/unsloth: 5X faster 60% less memory QLoRA finetuning
Python, C, Assembly - 2'500x Faster Cosine Similarity 📐 -- Ash's Blog
Prompt-Engineering-Guide/guides/prompts-adversarial.md at main · dair-ai/Prompt-Engineering-Guide
ReAct: Synergizing Reasoning and Acting in Language Models
How many CPU cores can you actually use in parallel?
VideoPoet – Google Research
Don’t starve, Diablo – Parallax 7 -- Simonschreibt.
A Lufthansa A350's Frustrating Oakland Diversion - One Mile at a Time
powerinfer-20231219.pdf
Making my own Bed Sensor — Home Automation Guy
Simulating Fluids, Fire, and Smoke in Real-Time
Build a search engine, not a vector DB
Paper page - LLM in a flash: Efficient Large Language Model Inference with Limited Memory
End of an era for electronics giant Toshiba - BBC News
Introducing llamafile - Mozilla Hacks - the Web developer blog: Llamafile is the easiest method of running models on your own hardware (if you use the "-server" variant). Simply wrap the model in a file/use a pre-wrapped model and execute it.
Zack Witten on X: "Thread of known ChatGPT jailbreaks. 1. Pretending to be evil https://t.co/qQlE5ycSWm" / X
Perplexity Labs
Victor Guyard -- Project - Creating a Disc Golf picker upper for Create-X
Colliding Secure Hashes -- Blog
Do I really need to get out the soldering-iron again? - Naughty Computer
Corvus (boarding device) - Wikipedia
LED Industrial Piercing - mitxela.com
UV Protection Amulet - mitxela.com
Mixtral-8x7B: Understanding and Running the Sparse Mixture of Experts -- by Benjamin Marie -- Dec, 2023 -- Towards Data Science: Mixtral 8x7B is a fantastic model (I use it on Perplexity) and this explanation is great.
Visualize Mixtral MoE: : Mixtral 8x7B is a fantastic model (I use it on Perplexity) and this exploration is great.
Paying Netflix $0.53/h, etc.: I've long used dollars per entertainment*hour as a metric for measuring various entertainments.
cognitivecomputations/dolphin-2_6-phi-2 · Hugging Face: The model is only a few days old, but already on HF.
etymology - Did English ever have a formal version of "you"? - English Language & Usage Stack Exchange
This strange vine can mimic other plants. How? - Vox

Weeklog for Week 51: December 18 to December 24

Progress

Articles

Libraries, programming, etc

Books

Games

Backlog

Other media