In the ever-expanding landscape of generative AI, Meta introduces Llama 2, a new family of AI models that competes directly with established players like OpenAI’s ChatGPT, Bing Chat, and Bard. Unlike its predecessor, Llama 2 is now freely available for both research and commercial use. Meta claims that Llama 2’s performance has significantly improved compared to its earlier version, thanks to its training on a mix of publicly available data and fine-tuning capabilities on platforms like AWS, Azure, and Hugging Face’s AI model hosting.
The enhanced accessibility of Llama 2 extends to its optimized compatibility with Windows and Smartphones, courtesy of an expanded partnership with Microsoft and Qualcomm. It offers two versions: Llama 2 and Llama 2-Chat, the latter being fine-tuned for two-way conversations. Both versions come in different sophistication levels: 7 billion, 13 billion, and 70 billion parameters. Meta’s researchers affirm that Llama 2 retains the architecture of its predecessor but is trained on 40 percent more tokens, enabling it to process larger input prompts from diverse text sources.
To improve the helpfulness and safety of its responses, the Llama 2-Chat model underwent fine-tuning with reinforcement learning, using human feedback techniques. While Llama 2 may not outperform the highest-profile closed-source rivals like GPT-4 and PaLM 2 in certain benchmarks, human evaluators rate it as roughly as “helpful” as ChatGPT, according to Meta.
Acknowledging that generative AI models possess biases, Meta emphasizes its commitment to addressing potentially harmful outcomes with Llama 2. Users must adhere to Meta’s license, acceptable use policy, and guidelines for safe development and deployment.
Despite some people claiming Llama 2 is open source, Meta clarifies its stance, promoting openness and innovation while being mindful of certain caveats. Llama 2 is not for use in illegal activities, deception, or misleading purposes. Additionally, developers cannot utilize Llama 2’s outputs to enhance other large language models without requesting a special license, especially if the applications support over 700 million monthly active users.
In conclusion, Meta’s Llama 2 joins the ever-growing generative AI landscape, presenting new possibilities for research and commercial use while striving to balance accessibility, performance, and safety within the realm of AI development.
