Patent Application
20240152723
This disclosure relates generally to an integrated human/AI interface and a system/AI interface and, more particularly, to an integrated human/AI interface and a system/AI interface, where an AI designed for a human and an AI designed for a system interact with each other so that human AI learns about the system and the system AI learns about the human.
Artificial intelligence (AI) employs interactive computer systems that perform a function or task that normally requires human intelligence, such as visual perception, speech recognition, decision-making, etc., and is known to be used as part of various systems to either assist humans or replace humans. Known AIs interface or interact with humans in different ways. Some AIs gather data on humans to be used for various predictive activities, for example, marketing, whereas other AIs are intended to be inherent to the particular system and the human is ancillary to that system. For most AI enhanced systems that interact with humans, there is a single instance/layer of AI interfacing with the human. Typically, these are system-oriented AIs that are not geared toward interfacing directly with the human. In other words, the AI is designed for the system and not the human. This has led to concerns of AIs and humans doing unexpected things during their interaction, ultimately not providing confidence for the user. For example, in the autonomous vehicle industry, there is a perception that if the AI systems that steer, accelerate and brake the vehicle would need a human in the vehicle to intervene in some situations, the human would adequately do that for those situations. However, the reality is that humans react differently to different things, and thus the confidence that the human will react in a certain desired way may not be reasonable. Therefore, one of the factors that will have a significant effect on the shifting of the human-system balance of task allocation in AI enhanced architectures is how well the system is able to create justified confidence (trust) in the way that it executes its responsibilities.
The following discussion discloses and describes an architecture that includes a human, a human AI agent designed to understand and interact with the human, a system, and a system AI agent designed to understand and interact with the system. The human AI agent and the system AI agent are configured to be in communication with each other in a manner so that the human AI agent learns about the system and the system AI agent learns about the human so as to optimize an interaction between the human and the system. The human AI agent and the system AI agent are configured so that the human AI agent learns about the system and the system AI agent learns about the human during a set up process before the architecture is put in operation and during the operation of the architecture. The human interacts with the system directly and through the human AI agent and the system AI agent and the system interacts with the human directly and through the system AI agent and the human AI agent.
Additional features of the disclosure will become apparent from the following description and appended claims, taken in conjunction with the accompanying drawings.
The following discussion of the embodiments of the disclosure directed to an integrated human/AI interface and a system/AI interface is merely exemplary in nature, and is in no way intended to limit the disclosure or its applications or uses.
This disclosure proposes an architecture that includes at least two AIs, where one of the AIs is designed and optimized for a human and one of the AIs is designed and optimized for a particular system. These two AIs are in communication with each other and through that communication each learns about the knowledge that the other has so that the AI for the human learns about the system and the AI for the system learns about the human. When it's time to bring the human and the system together, the two AIs then prepare an interface between the human and the system, ensuring that it is optimal for the specific task to be performed. The AIs would communicate through the same language or protocol, regardless of whether the human and system do, which creates certain efficiencies and make the system overall more effective. Rules between the AIs would be provided, such as who has priority at each decision point, how is the priority determined, what is the importance of each decision, etc. Because of the amount of data and length of exposure that it has had with the human, the human AI also understands patterns of the human when the human really understands a new system. Thus, when introducing a novel system, it can collaborate with the system AI to better identify whether the human is truly ready for operations than current training assessment methodologies. It is noted that the term “AI” as used herein could also be an autonomous agent, and the number of AIs and the number of actual agents (human and/or system) can be greater than two.
Designing the integration of the AIs as discussed above between a system and a human for a specific architecture starts with providing the interface 12 as shown in
Once the AIs 16 and 36 coordinate and optimize the interaction between the human 14 and the system 34 as a set up step, they continue to coordinate and optimize and cycle contingencies during the interaction between the human 14 and the system 24 while the particular architecture is in operation so as to optimize the architecture for different environments and scenarios. This is illustrated by architecture 50 in
The foregoing discussion discloses and describes merely exemplary embodiments of the present disclosure. One skilled in the art will readily recognize from such discussion and from the accompanying drawings and claims that various changes, modifications and variations can be made therein without departing from the spirit and scope of the disclosure as defined in the following claims.
