Meta announced today that it has launched a new language model for artificial intelligence called Llama, which stands for Large Language Meta Meta AI, to assist the scientific community in research in the field of machine learning and artificial intelligence.
Mark Zuckerberg, CEO of MITA, said in a Facebook post, “The llama model has shown impressive results in creating texts, conversations, written materials, and more complex tasks such as solving required math or protein structures. Inspiring results.”
The Llama model is an artificial intelligence algorithm that recognizes text, compiles it, translates it, and creates new content based on insights gained through massive data collection. Deep AI uses artificial neural networks to mimic the behavior of the human brain.
The problem that researchers in this field face is that large LLM language models such as chatgpt rely on a robust LLM computing infrastructure, which makes them unusable for most researchers. Meta says its new model will open the door for everyone to use big language models, not least because it is one of the most important forms of artificial intelligence.
The company has confirmed that it will make the Trinity model available under a noncommercial license, which means it will be limited to research use only. The model will be available in different sizes with 7 billion to 65 billion parameters.
Meta said it has trained its model on billions of databases, outperforming the JPT-3 model by only 300 million points. According to the company, this makes its model more versatile and able to provide broader support for custom GPT-3 tasks across a broader range of domains.
Finally, the company says it will provide undergraduate researchers with new models based on examining each research case and its meaning, and will focus on government and university research projects and civil society organizing projects.
The company did not say whether it plans to launch its own chatbot based on Chatgpt's Llama model or on Google Bard.