The LLAma Effect Just a few months ago, we only witnessed major players like Google, OpenAI, Microsoft, Amazon, and Meta making advancements in LLMs. Then something incredible happened, completely changing the game: LLama was leaked from Meta and triggered the research and the engineering community. 𝐋𝐋𝐚𝐦𝐚 was originally released in different formats: 7B, 13B, 33B, and 65B parameters, and, despite being smaller than GPT-3, it matched the performance in many tasks. So what? Well... The staggering amount of innovation triggered in just a few weeks is: ✅ Stanford released 𝐀𝐥𝐩𝐚𝐜𝐚, an instruction-following LLama model ✅ 𝐕𝐢𝐜𝐮𝐧𝐚 was released from University of California, Berkeley, Carnegie Mellon University, Stanford University, and UC San Diego. It's a fine-tuned version of LLama that matches 𝐆𝐏𝐓-4 performance ✅ Berkley AI Research Institute released 𝐊𝐨𝐚𝐥𝐚, a fined tuned version of LLama ✅ Nebuly open-sourced 𝐂𝐡𝐚𝐭𝐋𝐋𝐚𝐦𝐚, a framework for creating conversational assistants using your own data ✅ 𝐅𝐫𝐞𝐞𝐝𝐨𝐦𝐆𝐏𝐓 was released as an open-source conversational agent based on Alpaca ✅ UC Berkeley released 𝐂𝐨𝐥𝐨𝐬𝐬𝐚𝐥𝐂𝐡𝐚𝐭, a ChatGPT type model with RLHF pipeline based on LLama It is not finished here 🤯 ➡️ In addition, Databricks released 𝐃𝐨𝐥𝐥𝐲, the world's first truly open Instruction-Tuned LLM, which can be used for commercial uses 🎊 🎉 All of this happened in 40 days. Mind-blowing. Write your thoughts in the comments 🤞🏻 #ai #innovation #data #deeplearning #generativeai #deeplearningai #llama #chatgpt #research #engineering #google #university #microsoft #amazon #meta Via https://www.linkedin.com/in/nicola-massarenti