Meta Open-Sources LLaMA 2, to take on ChatGPT, and promote responsible AI innovation

Meta opens-sources LLaMA 2 in order to compete with ChatGPT

The more competition, the better. Download and spread the software around the globe so that large companies cannot seal off competition under the cloak AI safety.

Meta, formerly Facebook, announced recently that it will open-source its large language models (LLMs) called LLaMA 2 for free commercial and research usage. This is seen as an open challenge to OpenAI ChatGPT – the popular chatbot powered the GPT-4 Model, which isn’t open-sourced.

LLaMA 2 can generate natural language text based on an input or prompt. It can be used in a variety of applications, including chatbots and content creation. LLaMA 2 was released as the second version Meta’s LLM in February 2023. Meta claims that LLaMA 2 has been trained using 40% more data than LLaMA 1. This includes data from \”publicly accessible online data sources\”. Meta claims it also \”outperforms other LLMs such as Falcon and MPT\” when it comes reasoning, coding and proficiency tests.

Meta has decided to offer LLaMA 2 for free via Microsoft’s Azure platform as well as through other providers like AWS, Hugging face, and direct download. Meta stated that it wanted to provide businesses, researchers, and startups with more AI tools to allow for innovation and experimentation as a group. Meta said it was committed to \”building responsible\” with its AI system. The company stated that its models were tested for safety both internally and by externally using \”adversarial prompts\” to help fine-tune the model. Meta also reveals how models are evaluated and tuned.

Source:
https://medium.com/@SPX701/meta-open-sources-llama-2-to-compete-with-chatgpt-1370d587b104

发表回复

您的邮箱地址不会被公开。 必填项已用 * 标注