LLaMA Fever

Also in today's edition: Doomsayer Yudkowsky, BlackBerry Boys are Back & Inside BloombergGPT

Earlier this week, Hugging Face unveiled Vicuna-13B, a fine-tuned model based on Meta’s LLaMA. A few weeks ago, Stanford University released Alpaca, again, based on LLaMA. What’s interesting is that the latter was trained for just $600 (Rs 50,000) and Vicuna at an even lesser cost of ~ $300 (Rs 25,000).

The latest chatbot, built on Vicuna-13B, is said to have achieved 90% qualities of OpenAI’s ChatGPT as well as Bard, alongside outperforming other models like LLaMA and Alpaca in more than 90% cases. Check out the code here.