On July 18, 2023, Meta CEO Mark Zuckerberg said the company is affiliating with Microsoft. The collaboration will present the next generation of its AI large language model and develop the technology. The free-for-research and commercial technology is known as Llama 2.
Llama 2 is aimed at enabling developers and firms to construct generative AI-powered tools and better user experience. The company explained more details in a blog post. It said the firm has done away with the massive demand for Llama 1 from researchers. The firm received at least 100,000 requests for accessing the big language model.
The company said these are some of the attractive things the firm achieved to build the language. The firm is now prepared to open-source the next version of Llama 2. The new version is available free of charge for commercial and research purposes.
The Advanced Features in Llama 2
The Llama 2 has double the framework length of Llama 1 and is trained on 2 trillion tokens. The firm said the chat models of Llama 2 were also trained on more than one million new human observations. Azure customers can modify and easily set the 7B, 13B, and 70B-parameter Llama 2 models more securely on the platform.
A research paper from Meta said the new model depicts less openness compared to Meta. The platform has formerly shown a significant openness in its work to construct models. It needs the absorption of a big stock of digitized writings such as books, social media feeds, and news articles.
Meta’s paper said the latest version was prepared on a new blend of data from sources available publicly. However, it doesn’t include data from services or products of Meta without specifying which data was used. It said that Meta has erased data from websites found containing a higher volume of private individuals’ personal information.
Meanwhile, Meta used the acronym “LLaMA” (Large Language Model Meta AI). The firm mentioned the first version of its model and released it in February. The tech giant has now modified the capital letters and presented “Llama 2” as its 2nd version.
The Accessibility of Llama 2 Version
You can find Llama 2 in the Azure AI model catalog. It enables developers to use Microsoft Azure to develop and grasp their cloud-native tools for security and content filtering features.
The new version is also designed to run locally on Windows and gives developers an ideal workflow. Developers can bring reproductive AI experiences to customers through different platforms. You can get Llama 2 from Hugging Face, AWS (Amazon Web Services), and other providers.
How is Llama 2 Dissimilar from ChatGPT and Bard?
The new version of Llama 2 is free for commercial use and research purposes. Microsoft, Google, and ChatGPT are open-source but don’t provide similar offerings. The firm said Llama 2 defeats other open-source language models on many outside standards. It includes coding, reasoning, skillfulness, and knowledge tests.
The first version of Llama was formerly combative with models that power the Bard chatbot of Google and OpenAI’s ChatGPT. However, the new Llama was effectively trained on 40% more data compared to its ancestors. Zuckerberg said at least one million human observations are used to fine-tune the quality of its outcomes.
Meta has also attested to recognize it for becoming more open compared to its major tech competitors. The firm is offering a peek at the code and data it utilizes to develop AI systems. It stated that such exposure makes it simple for external researchers to help mention and mitigate intolerance and harmfulness.
The Impact Challenge of Llama 2 Version
AI systems pick up after observing how actual people write and communicate. On Tuesday, Zuckerberg said in a Facebook post that open-source motives inventions because it allows many more developers. It enables developers to construct with new technology. However, it enhances security and safety because the software is open.
More people are allowed to examine it to mention and fix possible issues. Zuckerberg said he believes it would uncover more advancement after opening more ecosystems. So, the company is open-sourcing Llama 2 version. Meta also said in its blog that it needs to motivate the community of inventors.
It is for those who desire to use Llama to solve critical issues. The company is launching a contest to motivate a diverse set of public, non-profit, and for-profit elements. They will use Llama 2 to address education, the environment, and other essential challenges. The company will provide contest rules before starting it.