- 3D LLM Visualization
3D LLM Visualization
PLUS: Upscaling images is getting MUCH better, Gemini (by Google) is being postponed
Hi folks!👋🏻 This is The Prompt!
Here’s what we have for you today:
Visualization on how LLM works
Gemini (by Google) is being postponed
OpenAI is buying AI chips
New idea to video platform shows incredible results
Upscaling images is getting MUCH better
Intro to multi modal RAG
Let's read on
Look into the LLM black box
Do you want to understand what’s happening inside an LLM?
This 3D visualization tool clearly demonstrates every stage of the process, showing what the LLM does at each step.
They’re using the model nano-gpt, with a mere 85,000 parameters. The task that they’re visualizing is sorting a sequence of six letters C B A B B C in alphabetical order i.e. to "ABBBCC".
You can then compare this small model to bigger ones like GPT-3 and see the difference.
🚨 What else is going on
Intel dropped a 7B model called NeuralChat fine-tuned on Mistral 7B
Google DeepMind researchers expanded the number of known stable materials on Earth by 380,000
Google postpones big AI launch for Gemini, apparently it didn’t handle non-English queries reliably
AI X Robotics is in full swing, there is a new framework for learning household tasks called Dobb-E
OpenAI agreed to buy $51M of AI chips from a startup backed by CEO Sam Altman
LogoDiffusion: Redesign any logo, with text prompts
Nova2: New Speech to Text API
DrawUI: Draw and build a website on your phone (code available)
PaperMark: Open source AI documentation assistant
Swifty: simplify business travel with your travel assistant
✍🏼 Prompt of the Day
A dense bamboo forest with pandas, East Asian forest, tranquil, natural habitat, gentle, immersive, bamboo green and earthy, 4K Ultra HD --style raw --ar 16:9
What'd you think of today's edition?