Google released PaLM 2 on May 10, 2023, as a valuable response to OpenAI GPT-4. At its recent I/O event, Google unveiled a fascinating lineup of PaLM 2 models, from the smallest to the largest: Gecko, Otter, Bison, and Unicorn. According to the Google PaLM 2 technical report (see Tables 5 and 7), PaLM2 is not only better, faster, and smaller than previous PaLMs, but also outperforms gpt-4 in some inference domains.
Like many others, here at Outside we are learning to use our LL.M. to better serve our outdoor community. Recently, we had the opportunity to test PaLM2 and GPT-3.5 using external real-world use cases. If you are considering choosing Google and OpenAI as your LLM provider, or you just want to learn how to build a Langchain agent equipped with a knowledge base tool for search and question answering, I hope this article can provide an insight into designing an evaluation suitable for your field frame.
In this post, I'll share our exploration of four key areas:
Methodology and technology platforms: Pinecone, Langchain, LLM (PaLM2 and GPT-3.5)
Inference speed and answer quality: Comparing the performance of Langchain's retrieval QA chain and conversational retrieval chain with code
examples react-description Proxy performance with Google Search API (SerpApi)
small talk and security issues
Side Note: The incantation I use to prompt midway to create a feature image is:
yellowstone park with rainbow background, vintage travel poster style, impressive landscape, impressive panoramas, — ar 16:9 — v 5