Meta, together with Microsoft, unveiled the open source artificial intelligence Llama 2

Meta has partnered with Microsoft to deliver Llama 2 AI as the next generation of language models for business and research. This model will be released as open source through Microsoft Azure, AWS, Hugging Face and other AI servers.

Microsoft unveiled new products such as Bing Chat Enterprise at the Inspire event. Meta also appeared at this event and announced a new artificial intelligence model in cooperation with the Redmond technology giant Llama 2 as open source offers As stated on Microsoft’s website, this model will be supported on Azure and Windows.

Additionally, Meta will partner with Qualcomm to launch Llama 2 via the Snapdragon platform starting in 2018. 2024 Bring it to phones, laptops and headsets. With this, there will be no need to rely on cloud services to use artificial intelligence capabilities in various apps.

Meta says he’s open sourced Llama 2 to safety This model will improve artificial intelligence. The company believes that by doing this, developers and researchers can test this large language model and find its problems.

Meta says that Llama 2 has also passed the testing process through the red teams. These tests use teams inside and outside the company to run aggressive tests against a product.

How is the artificial intelligence of Llama 2 different from Llama?

Llama 2 comes in two versions: Llama 2 And Llama 2-Chat. The second version is optimized for two-way conversations. The two versions of the second generation of this artificial intelligence are also three revisions in terms of the number of parameters 7, 13 and 70 billion parameters is divided

Llama 2 on 2 trillion tokens has been taught. Each token is a piece of text like a word. This is a significant increase compared to Llama’s 1.4 trillion tokens. Of course, among the competitors there are models like Google’s PaLM 2, which were trained on 3.6 trillion tokens.

Meta hasn’t revealed exactly what their source of information was for the Llama 2 tutorial, saying only that they got their information from the web. This AI model performs a little worse in benchmarks than models like GPT-4 and PaLM 2. But Meta says their model performed almost as well as ChatGPT in human evaluations.