A new AI research team at Cisco has collaborated with Meta to train its Llama 3 large language model specifically on cybersecurity data. This model will be open-sourced, including its weights, enabling widespread access for developers and researchers.
Key Points
- The Foundation AI team at Cisco is training the Llama 3 model on cybersecurity data.
- The model will be released as open source, allowing public inspection and fine-tuning.
- Engineers from Meta and Google contributed to the model’s training process.
- The project distilled 200 billion tokens of data to focus strictly on cybersecurity, resulting in a robust model.
- The LLM can perform efficiently on a single Nvidia A100 GPU, making it cost-effective for organisations.
Why should I read this?
If you’re at all interested in cybersecurity or AI development, this article is a must-read! The collaboration between Cisco, Google, and Meta has resulted in a powerful tool that could reshape how organisations handle security threats. This opens up new avenues for custom AI applications in tackling complex cybersecurity issues. So, sit back and let us fill you in on the details!