Generative AI Group Chat: Resources, Hackathons, and Model Discussions

Introduction

  • Group chat transcript on Generative AI
  • Chaotic discussion on various topics

Learning Resources

  • Link to a post on what transformers are
  • Cohere has two of the best ML content creators
  • List of top AI-themed newsletters shared
  • 42papers.com recommended
  • Ben’s Bites newsletter recommended

Hackathon

  • Invitation to join a team in Warpspeed GenAI hackathon

Deterministic Output

  • Discussion on making GPT2 or BLOOM model outputs consistent
  • Two ways to make it deterministic: setting temp=0 or setting seeds
  • Pro tip to pursue this line of reasoning and gain first-hand experience

Visual Aesthetic Scoring

  • Discussion on AVA dataset and image dataset curation methods
  • Link to a space on huggingface for aesthetic predictor
  • Use of clip interrogator to score generated captions
  • Hallucination and determinism are not related
  • Visual aesthetic scoring is largely a final layer problem
  • Discussion on the limitations of the model

Dataset Creation

  • Request for good library references to create datasets for LLMs
  • Discussion on creating a dataset based on few hand-written examples
  • Suggestion to check the terms of service for commercial use
  • Discussion on using a visual QA model to identify errors in a picture
  • Workflow suggestion to detect errors in the image and fix them using models finetuned on those datasets
  • Suggestion to use Andrew Ng’s landing.ai for the task
  • Discussion on using multiscale patches and CLIP to check similarity with the prompt

Finance GPT

  • Discussion on usable finance GPT
  • Bloomberg GPT is expensive
  • OpenBB is an open-source alternative to Bloomberg terminal

Autonomous Agents

  • Discussion on autonomous agents space
  • Link to a beginner’s guide to autonomous agents
  • GitHub repository for a simpler and no dependencies solution

The description and link can be mismatched because of extraction errors.

  • https://huggingface.co/spaces/mosaicml/mpt-7b-chat: This link leads to a post on Hugging Face’s website that explains what transformers are for beginners and those who want to deepen their understanding of the topic.
  • Check out 42papers.com for some interesting content. However, be aware that it may be similar to Twitter, with much of the content already coming from https://twitter.com/_akhaliq. Here are some useful links I’ve found:
  • https://huggingface.co/spaces/Geonmo/laion-aesthetic-predictor: Suggested as a tool to use for generating a score for a generated caption.
  • https://github.com/mosaicml/llm-foundry/tree/main/scripts/eval: The message in the same link as the URL mentions the need for a standard that everyone can use and expresses hope that HELM becomes that standard. The context of the URL is unclear.
  • https://landing.ai/ - suggested as a potential solution for a workflow involving multiple models for a task that may be difficult for a single model to handle.
  • https://llava.hliu.cc/ - Someone in the group was asking if anyone had experience finetuning GPT with Yahoo Finance. In response, someone shared a link to an open source alternative to Bloomberg terminal called OpenBB and mentioned that they released a blog on how to train on their documentation to get the appropriate OpenBB command.
  • https://www.mattprd.com/p/the-complete-beginners-guide-to-autonomous-agents: A beginner’s guide to autonomous agents that the speaker is starting to look into.
  • https://github.com/dosco/minds: The speaker mentions that they have built something similar to this link for themselves, which is much simpler and has no dependencies.
← Exploring AI Applications, Ethics, and OpenAI's Closed Source Approach
OpenAI, Google Models, and Hugging Face Developments →