Recently, Meta made waves by introducing their new artificial intelligence language model called Large Language Model Meta AI (LLaMA). The LLaMA model is a large language model created to help researchers in their work in the field of artificial intelligence (AI). This development has caused a stir in the tech world, with industry leaders scrambling to catch up with Meta's impressive achievement. Here’s the complete story.
The Story- Meta Unveils Advanced Language Model
As companies are struggling to incorporate AI technology into their products and win over investors, Meta Platforms Inc. made a significant announcement on February 24, 2023, to release a new large language model, which is the core software of a new AI system. This competition began with Microsoft-backed OpenAI's ChatGPT being released late last year, prompting other tech giants like Alphabet Inc. and China's Baidu Inc. to promote their own products.
Meta's advanced language model, called Large Language Model Meta AI (LLaMA), will allow researchers and organizations connected to the government, civil society, and academics to use the model under a non-commercial license. Large language models are capable of summarizing data and generating content by mining enormous quantities of text. They can respond to queries with sentences that appear to have been written by humans, making them a crucial component of AI technology.
Meta's LLaMA model is trained on 20 languages, with an emphasis on those with Latin and Cyrillic alphabets, and uses "far less" computing power than earlier solutions. Smaller models like LLaMA require less computing power and resources, making it easier for researchers who don't have access to large amounts of infrastructure to study them. This further democratizes access to AI, an important, fast-changing field. Unlike a fine-tuned model that is made for a particular job, LLaMA is designed to be flexible and can be applied to a variety of use cases.
Large language models, with billions of parameters, have shown new capabilities over the past year, but full research access to them remains limited due to the resources required to train and run such models. This has slowed down efforts to strengthen their robustness and lessen recognized problems like bias, toxicity, and the potential to spread false information.
Meta is making LLaMA available in several sizes (7B, 13B, 33B, and 65B parameters) and sharing a LLaMA model card that details how they built the model in keeping with their approach to responsible AI practices. LLaMA is trained on one trillion to 1.4 trillion tokens and focused on 20 languages with the most speakers, emphasizing those with Latin and Cyrillic alphabets.
Meta is disseminating the LLaMA model under a non-commercial license with an emphasis on research use cases in order to preserve its integrity and guard against exploitation. Academic researchers, people associated with organizations in government, civic society, and academia, as well as industry research laboratories around the globe, will be given access to the model on a case-by-case basis. Meta is excited to see what the AI community can discover and ultimately construct using LLaMA. The AI community, comprised of academic researchers, civil society, policymakers, and industry, must collaborate to create clear guidelines around responsible AI.
According to Gil Luria, senior software analyst at D.A. Davidson, Meta's announcement "seems to be a step in testing their generative AI capabilities so they can implement them into their products in the future."
Despite having less experience with generative AI, which is a new application of AI, Meta believes that it is an important area for the future of their business. With the tech sector's slowing growth leading to widespread layoffs and a reduction in experimental bets, AI has emerged as a promising area for investment.
Meta claims that LLaMA could outperform competitors who evaluate more parameters or variables. Specifically, they claimed that a version of LLaMA with 13 billion parameters could outperform generative pre-trained transformer (GPT-3), a more contemporary model on which ChatGPT is based. This announcement marks a significant step forward for Meta in the AI industry, and it will be exciting to see how their LLaMA model evolves and impacts the competition in the coming years.
Conclusion
Meta's latest language model, the LLaMA, has set a new benchmark in the AI race among big tech companies. The impressive capabilities of the model have garnered attention and caused industry leaders to take note. This development marks a significant step forward in the field of AI, and it will be exciting to see how it evolves in the coming years.
Interested to know more about the growing technologies in your industry vertical? Get the latest market studies and insights from BIS Research. Connect with us at [email protected] to learn and understand more.
