USA: Meta has entered the Generative AI competition along with Open AI's ChatGPT and Google's Bard. The LLaMA language model of business is fueling an AI arms race.
It has been promoted as a secure alternative to the AI chatbots of Google and the Microsoft-OpenAI alliance. To preserve its integrity and prevent misuse, LLaMA is made available under a non-commercial license.
When OpenAI's ChatGPT caught the world's attention last year, it triggered a contest for supremacy in the AI technology space that saw Alphabet Inc. and prompted Baidu to launch its own products.
Also Read: Netflix lowers its subscription costs by 60%
To ensure that businesses are not left behind in the market for supercharged generative AI, META, which has long been at the forefront of AI research and development, has now introduced LLAMA.
According to Meta, the Large Language Model Meta AI (LLAMA) is a "baseline, 65 billion parameter large language model". It was created to help researchers in the field of artificial intelligence and solve problems with language models for AI.
According to Meta's official blog, LLaMA is a "smaller, more performant" model that will enable research communities to study AI models without access to critical infrastructure.
Also Read: User will soon be able to report status updates on WhatsApp
Like other large language models, LLaMA builds text by predicting the next word from a set of input words. LLMs typically mine huge amounts of text to summarize information and prepare content.
But LLAMA from META uses "much less" computing power than earlier methods. It can outperform rivals that consider more parameters or variables.
LLaMA is a collection of 7B, 13B, 33B and 65B parameterized foundational language models. 1.4 trillion tokens were used to train the LLaMA 65B and 33B models, compared to one trillion tokens for the LLaMA 7B model. In particular, Open AI's GPT-3 (175B) can be outperformed by LLaMA 13B even though it is ten times smaller.
DeepMind and Google's Chinchilla70B and PaLM-540B models can be compared to the 65B LLaMA model.
The LLaMA will be made available under a non-commercial license to researchers and organizations affiliated with government, civil society, or academia.
According to Meta's blog, "We hope that making LLaMA available to the research community will accelerate the development of large language models, increase their robustness, and reduce known problems such as toxicity/bias."
Also Read: All One subscribers can now use Google's Magic Eraser feature
The business plans to release larger models in the future that were trained on the large pre-training dataset.