"an internet search engine that writes what it finds, but sometimes lies" is a pretty healthy mental model to have of ChatGPT to be honest (especially when you mix in Browse with Bing mode)
@simonw
-
LLaVA Vision Models Integration with LLM Tools
By
–
Yeah I'd absolutely consider that – I have some work going on at the moment to get LLaVA and other vision models working with my LLM tool, I can definitely see the two projects coming together in the future
-
Public Perception: Real AI Technology vs Science Fiction Concepts
By
–
I'm interested in both questions, but I'm particularly interested in whether people assume that technology labeled "AI" is similar to the science fiction idea of AI
-
Public Understanding of AI: Research on Mental Models
By
–
Has anyone seen any trustworthy research into what the general public thinks "Artificial Intelligence" actually means? Do many people assume "AI" systems are Jarvis/Data level? What kind of mental model do people outside of the tech bubble have here?
-
Reverse Engineering System Instructions for REST API
By
–
The REST API documentation doesn't cover system instructions yet, so I had to reverse engineer how to use them from the Python client library – notes on that here https://
github.com/simonw/llm-gem
ini/issues/6#issuecomment-2046460319
… The JSON looks like this: -
LLM-Gemini Plugin Update Adds System Prompt Support
By
–
New release of my llm-gemini plugin – it can now send system prompts to the Gemini model, taking advantage of a feature they released today https://
github.com/simonw/llm-gem
ini/releases/tag/0.1a2
… -
GPT-4 Vision JSON parsing and prompt engineering techniques
By
–
Were you asking GPT-4 Vision for JSON and then parsing what came back – did you have to do anything special to avoid "Here is the JSON for that:" prefixes and suchlike?
-
Calling the New Embedding Model REST API Documentation
By
–
I'm having a tiny bit of trouble figuring out how to call the new embedding model – is there any clear REST API documentation anywhere?
-
Mistral Releases 8x22B MoE Model Advancing Open Source AI
By
–
Apparently this is a 8x22B MoE (Mixture of Experts) model Mistral's previous most powerful openly licensed model was Mixtral, an 8x7B MoE