SynCog: Advanced Artificial Intelligence for Autonomous Systems
Challenge
There are two types of problems: determinate and indeterminate. Current Artificial Intelligence (AI) technologies have focused on unconditional, unambiguous, determinate problems such as classification, pattern recognition, and data mining. Current AI technologies can only deal with problems they’re been trained to solve but fail when a problem falls outside of their domain knowledge. Because they are limited to unambiguous problems, current AI technologies are prone to fail in dynamic, ambiguous, or unpredicted situations.
Solution
ARA’s SynCog addresses an unmet market need. SynCog is leading-edge artificial intelligence designed for complex missions in uncertain environments. SynCog enabled systems feature situational awareness, responsiveness, fitness, and flexibility.
Biological brains have evolved to solve both determinate and indeterminate problems using a process called cognition. SynCog is the digital version of this process; it is a flexible “brain” for the product/system it autonomously controls.
SynCog-based systems expect dynamic change, continually learn, and improve performance throughout their lifecycle. SynCog autonomously learns the way animals learn, without continuous supervision. It is technology-independent and can be implemented in software and/or hardware. SynCog can be scaled to the complexity of its mission.
Unlike current AI, cognition is not trained from databases – instead it learns from mistakes and unexpected events. It observes interactions and constructs an internal model of the world based on those experiences. It learns by constantly comparing its model to the outside world and correcting errors.
SynCog enabled systems:
- Efficiently fuse multiple sensors and data sources in real time to perform complex, dynamic missions and tasks
- Can collaborate with humans and SynCog-enabled systems
- Train in both simulations and real-world environments
- Demonstrate causal reasoning, flexibility, efficiency, and interpretability
- Learn from experience, and improve performance throughout their lifecycle
