META’s Llama Might be Better than GPT-4 Itself.

Stay at least 30 seconds to see if you like this stuff

Meta has leapfrogged itself to the top of the AI competitors.

Meta has shocked the tech world with the release of Code Llama, an open-source large language model (LLM) specialized for coding.

I am fearful for its implications on Software engineers, which is something everyone seems to be doing, because of their high pay.

Code Llama incldues has three versions in its release: the foundational 70 billion parameter model, a Python-specific version fine-tuned for Python coding, and Code Llama Instruct, which is optimized for understanding natural language instructions.

In August 2023, Meta introduced the initial version of Code Llama as a state-of-the-art AI for coding. As of January 2024, they have enhanced this offering with the 70 billion parameter version

The Scary Part

All versions are free for both research and commerical use under Meta’s community license…

Meta’s Llama is extremely vast, with the three models catering to three completely different needs:

• The foundational model is a general text completion engine
• The Python model is specialized for Python programming