A New Scientific Achievement in Artificial Intelligence and Safety

A team of scientists headed by Radics Verification Department manager Oleg Odarushchenko, has been granted a patent for a utility model titled: "Method for Reserving an Artificial Intelligence System."

What is the essence of the invention?

The proposed method improves the reliability and safety of artificial intelligence (AI) systems by employing multiple parallel AI channels. The unique feature is that these channels utilize different AI models, trained with diverse tools and datasets. This approach reduces the likelihood of simultaneous failure caused by shared model defects and ensures better system behavior control in non-standard situations.

How does it work?

Performance Monitoring: All AI channels are evaluated for operability using specialized tools, which include:

• Input Data Analysis: The system assesses the input data of each channel, identifying whether the data align with or deviate from areas of uncertain or potentially hazardous behavior.
• Priority Setting: If there are operable channels with safe input data, the system selects data from one of them based on a predefined priority.
• Safety Signals: If all input data from the channels fall into a risk zone, the system generates a signal indicating uncertainty or danger.

What are the advantages of this method?

• Enhanced Safety: By identifying and accounting for zones of hazardous behavior, the method significantly reduces the risks of failure.
• Increased Reliability: The system can adaptively reconfigure itself and remain operational even when defects affect multiple channels.
• Improved Trustworthiness: Monitoring input data and uncertainty zones ensures higher quality and reliability of the output information.

This invention represents a major milestone in artificial intelligence, integrating innovative solutions to ensure the safety and reliability of critical systems.

What does this mean for the future?

From healthcare and transportation to energy and aviation, this reservation method has the potential to create intelligent systems capable of maintaining safety and reliability even under the most challenging conditions.