The year 2023 has been a year of unprecedented development in the field of Artificial intelligence. After the release of ChatGPT and its integration with Microsoft Teams and Bing, the environment in the AI arena is heating up. To counter this one-sided inclination Google came up with its own AI model Bard which resulted in around 12000 layoffs. Now the third contender Meta joins the AI race and releases a new language model LLaMA.
LLaMA or Large Language Model Meta AI, is a large language model that is said to assist researchers to advance the work in other verticals of AI. It is not just a single entity but with various sub-versions that are much smaller in size. To summarize it is “a collection of foundation language models ranging from 7B to 65B parameters.”
- The idea behind LLaMA is that smaller language models will hugely magnify the speed of research in AI as a large community does not have access to the infrastructure needed to study these models.
- This launch does not really show its intent in competing with other AI models like ChatGPT or Bard that have the ability to communicate.
Meta also provided an explanation in lucid form why they chose smaller models. It is because they are easy to retrain as they are trained on pieces of words or tokens. It also provided that its sub-models LLaMA 65B and LLaMA 33B are trained on 1.4 trillion tokens while the smallest model LLaMA 7B is on one trillion tokens.
Although it will not be available commercially right now for reasons stated in the blog. “To maintain the integrity and prevent misuse, we are releasing our model under a noncommercial license focused on research use cases. Access to the model will be granted on a case-by-case basis to academic researchers; those affiliated with organizations in government, civil society, and academia; and industry research laboratories around the world.”
In Meta Research paper, the abstract stated that,”In particular, LLaMA-13B outperforms GPT-3 (175B) on most benchmarks, and LLaMA-65B is competitive with the best models Chinchilla70B and PaLM-540B.” These best Models belong to DeepMind and Google respectively. Also, it revealed that the company only makes use of publicly available data which makes the model compatible with open-sourcing. This is unlike AI models such as DeepMind Chinchilla, Google PaLM, or OpenAI GPT-3 which use undocumented or publicly unavailable data.
Recommended: How to Integrate ChatGPT into Microsoft Word
After meta joins the AI race and releases a new language Model LLaMA, has been at odds with Google Bard release with a little buzz. This is not something bad as it is not rushing the release that resulted in Google Bard giving wrong response and a lot of criticism. This is something similar to Bing Chatbot that revealed too much about its system like its actual name is Sydney. It also resulted in Microsoft restricting Bing AI chat to 5 replies.
The announcement of LLaMA was also quite one with just a Facebook post and a message on Instagram from Meta Channel. It stated,” Today we’re going to release a new state-of-the-art AI large language model called LLaMA designed to help researchers advance their work. LLMs have shown a lot of promise in generating text, having conversations, summarizing written material, and more complicated tasks like solving math theorems or predicting protein structures. Meta is committed to this open model of research and we’ll make our new model available to the AI research community.” Meta is also excited about how the community will learn and build using LLaMA.
Stay tuned for more updates on tech and the latest innovations in AI on TechCult!
Source: Meta AI Blog