Links about AI - first level curation

"Role prompting"... telling the model to assume a role has never been a good way to elicit capabilities/style/etc. For instance, if you ask one of the Claude models to simulate Bing Sydney, assuming you can get it to consent, the simulation will probably be very inaccurate. But if you use a prompt that tricks them into... See more

One line, one word. ✏️🔤 A simple #p5js sketch where you draw a line and Gemini tries to describe it in a single word. I find it really fun to have the model write on the canvas with me (wherever drawing finishes). Uses Gemini 1.5 Flash API. Try it + fork/remix my sketch here: https://t.co/Ks5qsmYNH6

The first rule of machine learning: Do not start with machine learning. Unfortunately, to a hammer, everything looks like a nail. Whenever I get a call from a company to help them, they have already made their mind up: they want to use machine learning. It's... See more

Create OpenAI like AI assistant with Llama-3 deployed locally on your computer (100% free and without internet): https://t.co/mLIOGM6Ly1

Why Chat With PDF Is Hard And How ChatLLM Gets It Right Chatting on long docs is hard because most LLMs other than Gemini don't have a large context. However, even with Gemini's 1M context length, in-context learning is hard, and if you stuff the doc in the context, it doesn't do a good job.... See more