Meta has made a significant move in the field of AI development by releasing its latest Llama 2 large language model. This model has performed exceptionally in tests, outperforming other AI chat models, including OpenAI's GPT, in most benchmarks. The benchmarks include key parameters like helpfulness and safety.
Meta will offer Llama 2 free of charge for commercial use, giving an alternative to the existing large language models provided by Google and OpenAI. This could position Meta as a forerunner in the burgeoning AI development arena.
The company will release three versions of the model, trained on 7 billion parameters, 13 billion parameters, and a massive 70 billion parameters. Along with these, Meta is also launching 'Llama 2 Chat', a model fine-tuned explicitly for conversational applications.
An exciting development is the expansion of Meta's partnership with Microsoft. This will enable developers using Microsoft tools to choose between Meta's Llama and OpenAI's GPT models when building their AI experiences. This collaboration makes Microsoft a vital platform, facilitating the connection between users and these top LLMs.
One of the standout features of Meta's Llama 2 model is its focus on safety and ensuring the accuracy of the results it produces. To achieve this, Meta has invested in training around elements like 'truthfulness', 'toxicity', and 'bias'. As per Meta, Llama 2 Chat demonstrates significant improvement in terms of truthfulness and toxicity over the pre-trained Llama 2.
All these advancements make Llama 2 a potentially reliable and safe tool for generative AI, which could significantly broaden its use cases.
Microsoft Azure AI customers will have the opportunity to test Llama 2 using their sample data, providing real-world insights into its performance.
You can read more about the Llama 2 process and dataset here.