LLAMA - SCI & TECH
News: What is Llama 3, Meta’s most
sophisticated and capable large language model yet?
What's in the news?
●
Meta Platforms, the parent company of Facebook, Instagram, and
WhatsApp, has introduced its most capable Large Language Model (LLM), the Meta
Llama 3.
Llama 3:
●
It stands for Large
Language Model Meta AI. This is a family of LLMs introduced by Meta AI in
February 2023.
Versions of Llama:
●
The first version of the model was released in four sizes — 7B, 13B, 33B, and 65 billion
parameters.
●
According to reports, the 13B model of Llama
performed better than OpenAI’s GPT-3, despite GPT-3 having 135 billion
parameters.
Features:
●
It’s said to be the most advanced model, showing big improvements in performance and AI abilities.
●
It is based on the Llama 2 architecture which has
been released in two sizes, 8B and 70B parameters.
○
Both sizes include a base model and a fine-tuned
version that is specifically designed to improve performance on certain tasks.
●
The instruction-tuned version is made for enhancing AI chatbots that are designed
to engage in conversations with users.
●
All Llama 3 models handle context lengths of up to 8,000 tokens.
○
Having more tokens means users can input more
content or prompts, and the model can respond with more content.
○
This enables more interactions and better handling of complex inputs compared to Llama 2
or 1.
Languages:
●
At present, Meta AI is available in English across the US on WhatsApp.
Significance:
●
It can be used
on Facebook, Instagram, WhatsApp, Messenger, and the web.
●
It’s easily accessible for developers since Meta
has integrated the LLM into the Hugging
Face ecosystem.
●
It has the ability to engage in conversations and
respond to posts directly on the Facebook feed.