Wondering if GPT builder can change the category on his own. I never use it tbh cuz it always messes up the prompt
LLMS
-
Using regex and LLMs to detect spam in timelines
By
–
My point was that most spam in my timeline could be caught with a regular expression, and difficult cases with an LLM
-
Updates to Custom GPT Store and Chat Interface Flow
By
–
This also comes with a new flow of opening custom GPTs from the store where you open an about page first and one extra click will be needed to start a chat. On the About page, you see conversation starters and aggregated ratings are expected to be added later.
-
Model Comparisons for Testing and Results Validation
By
–
Love this. These model comparisons are perfect for double checking my tests/results to
-
Text versus 32-bit floats: lossless compression efficiency trade-offs
By
–
I don’t think that’s necessarily true! text is actually relatively few bytes, while each 32-bit float is 4 bytes, so at least at short lengths things could be lossless
-
Custom GPTs to gain personalized memory access features
By
–
WIP: Custom GPTs will have access to personalised memory in the future. There is a new flag in the GPT builder UI allowing builders to enable memory access. Currently, it breaks GPT when enabled. GPT memory is new Cookies
-
Text Embeddings Lossy Collisions Detection Problem
By
–
yeah! but I don’t know if text embeddings are necessarily lossy? and if so how to locate collisions? (but yes; this is the entire point)
-
VR Girlfriend Technology: LLM and AR Integration for Real-World Second Life
By
–
VR girlfriend is there. Insane, LLm+AR= realworld second life
-
Analyzing AI search index and knowledge cutoff limitations
By
–
Does it have access to the fresh search index? Feels like it provides outdated info on niche topics Or is it a cutoff effect?
-
User experience with Gemini and GPT capabilities
By
–
This one is probably the most interesting cuz Gemini looks ahead. But! A lack of API or bots makes it way less useful. I am still trying to teach GPT not to write sentences like the 1st one and keep failing