"Falcon LLM: The New King of Open-Source Large Language Models"

2023-06-08

Falcon LLM, a new large language model, has taken the crown from LLaMA. We see new large language models emerging every week, with an increasing number of chatbots available for our use. However, it is difficult to determine which one is the best and which one's progress is the most useful. HuggingFace has an open leaderboard for large language models, providing real-time tracking, evaluation, and ranking. They use a special framework to test generative language models on different evaluation tasks. LLaMA has long been at the top of the leaderboard but has recently been replaced by a new pre-trained large language model called Falcon-40B. About the Technology Innovation Institute Falcon LLM was created and built by the Technology Innovation Institute (TII), which is part of the Advanced Technology Research Council of the Abu Dhabi government. The government is responsible for technology research across the United Arab Emirates, and the team of scientists, researchers, and engineers is dedicated to achieving transformative technology and discoveries in the field of science. What is Falcon-40B? Falcon-40B is a foundational LLM model with 40 billion parameters trained on 1 trillion tokens. It is an autoregressive decoder model, meaning it is trained to predict the next token in a sequence given the previous token. The GPT model is a good example of this type of model. Falcon's architecture has been proven to be significantly better than GPT-3, requiring only 75% of its training compute budget for inference. Ensuring high-quality data at scale is an important focus for the Technology Innovation Institute team, as they understand that large language models are highly sensitive to the quality of training data. The team has built a data pipeline that can scale to tens of thousands of CPU cores for fast processing and is capable of extracting high-quality content from the web through techniques like duplicate data filtering and removal. In addition, they have smaller versions available: Falcon-7B with 7 billion parameters. If you are looking for a chat model that is readily available, they also offer Falcon-40B-Instruct and Falcon-7B-Instruct models. What can Falcon-40B do? Similar to other large language models, Falcon-40B can: - Generate creative content - Solve complex problems - Support customer service operations - Serve as a virtual assistant - Perform language translation - Conduct sentiment analysis - Reduce and automate "repetitive" tasks Falcon LLM is open source They have made the source code of Falcon LLM publicly available, making it easier for researchers and developers to use Falcon-40B and Falcon-7B. Large language models used to be limited to research and commercial purposes, but now they have become open source to meet the global demand for artificial intelligence. This is because the United Arab Emirates is highly focused on the significant role artificial intelligence will play in the future. Apache 2.0 aims to foster a collaborative, innovative, and knowledge-sharing ecosystem in the field of artificial intelligence, ensuring the security and safety of open source software. In summary, as the best open source model currently available, Falcon LLM impresses with its powerful optimization architecture, unique open source license, and two parameter sizes: 40B and 7B. Source: https://www.kdnuggets.com/2023/06/falcon-llm-new-king-llms.html

RECENT AI NEWS

RECENT AI TOOLS