Jian Zhang explores advanced AI Function Calling techniques for enhanced Cybersecurity and beyond.
Subscribe: Apple • Spotify • Overcast • Google • AntennaPod • Podcast Addict • Amazon • RSS.
Jian Zhang is co-founder, CTO, VP Engineering at Nexusflow AI a startup that uses Generative AI to build tools for Cybersecurity. This conversation revolves around the integration of various AI components, with a specific focus on cybersecurity and function calling copilots. It delves into the challenges of data curation and benchmarking, the importance of measuring function quality and capability, and the concept of a co-pilot for function calling.
Shortly before we recorded this episode, Nexusflow announced the release of NexusRaven-V2, NexusRaven is an open source 13 billion parameter language model that achieves state-of-the-art performance on zero-shot function calling.
Function calling co-pilots refer to AI systems that can interact with and use external tools and APIs based on user instructions, often given in natural language. This allows the AI to execute specific functions or tasks by calling on resources beyond its own internal capabilities, enabling more interactive and versatile uses of AI. For example, in software development, such an AI could automate or assist coding tasks by integrating with and utilizing various development tools and libraries. The key ideas are the AI’s abilities to:
- Interact with and utilize external tools/APIs
- Execute specific functions/tasks based on natural language user instructions
- Perform complex actions beyond its built-in capabilities
- Automate tasks by calling on external resources as needed
- Assist users in areas like software development by integrating with relevant tools/libraries
NexusRaven-V2 outperforms models like GPT-4 on real-world function calling benchmarks without any task-specific fine-tuning. NexusRaven was trained on open source data and can be freely used, even commercially. It comes with utilities to integrate it into software workflows as a drop-in replacement for proprietary function calling APIs. The creators have also open sourced benchmark datasets and a leaderboard to track progress on this key capability.
Interview highlights – key sections from the video version:
- Nexus Raven: what is it, and a deep dive into its key features
- Function Calling and Tool Usage in AI Models
- Connection with retrieval augmented generation
- Building Nexus Raven
- Function Calling co-pilots
- Relevant benchmarks
- Nexus Raven use cases and usage patterns
- Nexs Raven’s nearterm roadmap
- Generative AI and cybersecurity
- Generative AI in the hands of attackers
- Python software supply chain vulnerabilities
Related content:
- A video version of this conversation is available on our YouTube channel.
- Casey Ellis: The Future of Cybersecurity – Generative AI and its Implications
- Philipp Moritz and Goku Mohandas: Navigating the Nuances of Retrieval Augmented Generation
- Waleed Kadous: Best Practices for Building LLM-Backed Applications
- Malte Pietsch: Orchestration for LLM and RAG applications
- What We Can Learn from the FTC’s OpenAI Probe
- Large Language Models in Cybersecurity
- Entity Resolution: Insights and Implications for AI Applications
- Ram Shankar: Securing machine learning applications
- Dan Geer and Andrew Burt: Security and privacy for the disoriented
- Open Source Principles in Foundation Models
- Michele Catasta: Software Development with AI and LLMs
If you enjoyed this episode, please support our work by encouraging your friends and colleagues to subscribe to our newsletter:
