Key Technical Milestone in Causal Neurosymbolic AI Development
//
In the past few months we have achieved a significant milestone in demonstrating the benefits of a causal neurosymbolic architecture to power AI agents and co-pilots.
Proof of Value. This architecture is a groundbreaking fusion of causal reasoning and neurosymbolic AI, which is essential for creating fully interpretable, intelligent systems capable of understanding and reasoning about the world in a way that is similar to human cognition. This enables our AI to make predictions and decisions that are not only data-driven but also grounded in an understanding of cause and effect.
Demonstration of an Offensive Security Co-Pilot. Our first demonstration of this technology is an AI co-pilot for training penetration testers and red teamers. Our AI co-pilot assists junior and more seasoned pen testers by providing recommendations on what to do next and why. The ability to understand and manipulate causal relationships in a cyber environment allows the AI co-pilot to evaluate what the next best action should be given the intended goals of the penetration test. For example, a test that is fast and exhaustive will be noisier than a test that is seeking to remain hidden and undetected. Because our AI co-pilot understands that certain exploits can generate more noise on the network and are subsequently scored low (undesirable) if the pen test is prioritising stealthiness.
We are continuing to work with our customers on extending and enhancing the functionality of our Offensive Security co-pilot. We will be making more announcements about our first product offering and how qualifying organisations can get access in the coming months.
Comparative Testing. Our provisional testing shows that our causal neurosymbolic approach is superior to Large Language Models and Reinforcement Learning techniques for the pen testing use-case. Make sure to check our Journal where we will be providing more information on this shortly.
What is Causal Neurosymbolic AI?
Our AI architecture integrates three approaches to artificial intelligence:
Causal: an understanding of cause-and-effect in the intended work environment
Neuro: deep pattern recognition, classification and prediction using neural networks (aka. deep learning) that continually learns from data
Symbolic: structured knowledge and logic to ground the operation of the AI in highly reliable knowledge sources