Yeah, it has lots of issues now. Storing task/result pairs in Pinecone, and pulling relevant stuff for execution – but will be much easier to do complex stuff with 32k token limit.
LLMS
-
HELM Benchmarks Don’t Predict Real World Model Performance
By
–
academic tasks, doesn't reflect real world use. models scoring well on helm doesn't necessary mean they are good (or vice versa).
-
Ghostwriter Chat: Conversational AI Integrated Into IDE
By
–
A month ago, we released Ghostwriter Chat. Conversational AI directly in the IDE complete with a debugger and file context. We wrote a blog on how we built it, including technical challenges and prompt construction. https://
blog.replit.com/ghostwriter-bu
ilding
… -
Debugging Pinecone metadata retrieval and task deduction issues
By
–
Yeah, it’s being stored in Pinecone and I’m pulling 5 most relevant – but need to debug pulling out metadata as all it’s getting is the similarity score now (not helpful). That should fix repeating. Also deducing tasks isn’t done by OpenAI yet – so need to fix that.
-
LangChain Task Execution with Separate OpenAI Calls
By
–
Oh yeah, the thought/observation is using @langchain
, which I'm using to execute each tasks. Separate OpenAI call for task creation/reprioritization -
User Joins GPT-4 API Waitlist After Seeing Others Access
By
–
Yep – I just forgot to add myself to the gpt4 api waitlist until I saw others getting it…
-
Systematic Blind Evaluation Benchmarks for LLMs Urgently Needed
By
–
friendly reminder to everyone that there isn't yet a good & proper systematic blind eval/benchmark of LLMs yet, especially those on real world data/use-cases. if i were in academia this is something i'll work on immediately.
-
Upgrading from GPT-3 to GPT-4 with API Access
By
–
It’s gpt3, but I’ll upgrade to 4 when I get api access
-
Google Launches BARD AI Language Model for Natural Language Processing
By
–
Say hello to Google's new AI language model BARD which is here to shake things up! With early access to this innovative model, natural language processing and generation possibilities are endless. Discover more here: https://
bit.ly/3JPu9OA @GoogleAI @OpenAI @sundarpichai @bing -
Baidu Cancels Public Launch of Ernie LLM for Private Testing
By
–
Baidu’s Ernie LLM public launch cancelled in favor of more private testing