Human-System AI
An architecture with dual AIs for humans and systems optimizes interactions by mutual learning, addressing unpredictable AI-human interactions and enhancing trustworthiness and efficiency.
Patent Information
- Application Number
- JP2025526682
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2022-11-09
- Filing Date
- 2023-11-06
- Publication Date
- 2025-10-24
AI Technical Summary
Existing AI systems are designed for either systems or humans but not both, leading to unpredictable interactions and a lack of trustworthiness in AI-augmented systems, particularly in scenarios requiring human intervention.
An architecture featuring two AIs, one tailored for humans and one for systems, that communicate and learn about each other to optimize interactions, ensuring mutual understanding and trustworthiness.
Enhances the effectiveness of human-system interactions by providing accurate assessment of human readiness and optimizing system performance through iterative learning and adjustment.
Smart Images
Figure 2025535598000001_ABST
Abstract
Description
[Technical Field]
[0001]
[0001] This disclosure relates generally to integrated human / AI interfaces and system / AI interfaces, and more particularly to integrated human / AI interfaces and system / AI interfaces in which an AI designed for humans and an AI designed for a system interact with each other so that the human AI learns about the system and the system AI learns about the human.
[0002] Artificial intelligence (AI) employs interactive computer systems to perform functions or tasks that typically require human intelligence, such as visual perception, speech recognition, and decision-making. This function or task is known to be used as part of various systems to assist or replace humans. Known AIs interface with or interact with humans in various ways. Some AIs collect data about humans for use in various predictive activities, such as marketing, while others are intended to reside within specific systems, rendering humans subordinate to them. In most AI-augmented systems that interact with humans, it is one instance / layer of the AI that interfaces with humans. Typically, these are system-oriented AIs that are not designed to interface directly with humans. In other words, the AI is designed for the system, not for humans. This leads to concerns that AIs and humans may behave unexpectedly during their interactions and ultimately be untrustworthy by users. For example, in the autonomous vehicle industry, if an AI system steers, accelerates, and brakes a vehicle but requires a human in the vehicle to intervene in certain situations, it is assumed that this human will do so appropriately in these situations. However, the reality is that humans respond differently to different things, and therefore confidence that a human will respond in a particular desired way may be unjustified. Therefore, one of the factors that critically influences the shifting balance of human-system task allocation in AI-augmented architectures is how well the system can generate justified trustworthiness (trust) in fulfilling its responsibilities. Summary of the Invention [Means for solving the problem]
[0003]
[0003] The following discussion discloses and describes an architecture including 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 communicate with each other, with the human AI agent learning about the system and the system AI agent learning about the human, to optimize the 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 configuration process before the architecture is put into operation and during 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.
[0004]
[0004] Further features of the present disclosure will become apparent from the following description and appended claims, taken in conjunction with the accompanying drawings. [Brief explanation of the drawings]
[0005] [Figure 1] FIG. 1 is a block diagram of a known architecture including human / AI interfaces and systems. [Figure 2] FIG. 2 is a block diagram of a known architecture including a system / AI interface and a human. [Figure 3] Figure 3 is a block diagram of the human / AI interface. [Figure 4] Figure 4 is a block diagram of the system / AI interface. [Figure 5] Figure 5 is a block diagram of the architecture including the human / AI interface and the system / AI interface in the architecture setting step. [Figure 6] Figure 6 is a block diagram of the architecture, including the human / AI interface and the system / AI interface, in operation. DETAILED DESCRIPTION OF THE INVENTION
[0006]
[0011] The following discussion of embodiments of the present disclosure, directed to integrated human / AI interfaces and system / AI interfaces, is merely exemplary in nature and is in no way intended to limit the present disclosure or its application or uses.
[0007]
[0012] 1 is a block diagram of a known architecture 10. Architecture 10 includes a human / AI interface 12 having a human 14 and an AI 16, where AI 16 is specifically designed and implemented to know about and understand human 14. Architecture 10 also includes a system 18. System 18 interacts with human 14 either directly, as represented by line 20, or through interface 12, as represented by line 22, but AI 16 is not specifically designed to know about or understand system 18. System 18 is intended to represent any system capable of adapting to human interaction that is augmented by some form of AI.
[0008]
[0013] 2 is a block diagram of a known architecture 30. Architecture 30 includes a system / AI interface 32 having a system 34 and an AI 36, where the AI 36 is specifically designed and implemented to know about and understand the system 34. Architecture 30 also includes a human 38. The human 38 interacts with the system 34 either directly, as represented by line 40, or through an interface 32, as represented by line 42, but the AI 36 is not specifically designed to know about or understand the human 38. System 34 is intended to represent any system augmented by some form of AI and capable of interacting with humans.
[0009]
[0014] Figures 1 and 2 illustrate the problem discussed earlier, where known AI is designed for specific system or human capabilities, but not both. Significant effort is expended in designing AI to match and ideally enhance human use of the system, but expectations are often not met. Conversely, when these same systems are designed for humans, significant time and money is also expended selecting and training humans to use these systems.
[0010]
[0015] This disclosure proposes an architecture that includes at least two AIs. One of these AIs is designed and optimized for humans, and the other is designed and optimized for a specific system. These two AIs communicate with each other, and through this communication, each learns about the other's knowledge, so that the human-tailored AI learns about the system and the system-tailored AI learns about humans. When the time comes to introduce the human and the system, the two AIs prepare the interface between the human and the system and ensure that they are optimally suited to perform the specific task. These AIs communicate through the same language or protocol, regardless of whether the human and the system perform specific efficiencies or enhance the overall effectiveness of the system. Rules may be established between these AIs, such as which one to prioritize at each decision point, how to determine the priority, and the importance of each decision. Due to the amount of data and length of exposure the AI has had with humans, when humans actually understand a new system, the human AI also understands human patterns. In this way, when introducing a new system, the human AI can work with the system AI to identify whether the human is truly ready for operation, more accurately than current training and evaluation methods. It should be noted that the term "AI" as used herein can also refer to autonomous agents, and the number of AIs, and therefore the number of actual agents (humans and / or systems), can be greater than two.
[0011]
[0016] As discussed above, designing the integration of AI between a system and a human for a particular architecture begins with providing an interface 12 including a human 14 and an AI 16 specifically designed for the human 14, as shown in FIG. 3, and then providing an interface 32 including a system 34 and an AI 36 specifically designed for the system 34, as shown in FIG. 4. As a configuration step for the final architecture, the two AIs 16 and 36 are placed in communication with each other, as illustrated by architecture 44 shown in FIG. 5. The AI 16 develops a detailed understanding of the human 14 and can provide all of that information to the AI 36, and the AI 36 develops a detailed understanding of the system 34 and can provide all of that information to the AI 16. The AIs 16 and 36 exchange and analyze information so that they can together identify the capabilities and limitations of the human 14 and the system 34 and optimize the interaction between the human 14 and the system 34 to achieve the desired outcome. The interaction between the human 14 and the system 34 during the configuration step is represented by line 46 and can also be used by the AIs 16 and 36 to achieve the desired outcome.
[0012]
[0017] Once the AIs 16 and 36 have adjusted and optimized the interaction between the human 14 and the system 34 as a setup step, they continue to iteratively adjust and optimize contingencies between the interaction between the human 14 and the system 24 while the particular architecture is operational, optimizing the architecture for different environments and scenarios. This is exemplified by architecture 50 in FIG. 6, where the human 14, AI 16, system 34, and AI 36 are combined into one interface 52.
[0013]
[0018] The above discussion merely illustrates an example of the present disclosure. Those skilled in the art will readily appreciate that various changes, modifications, and variations can be made from the above discussion and the accompanying drawings and claims without departing from the spirit and scope of the present disclosure.
Claims
1. The architecture is Humans and a human artificial intelligence (AI) agent designed to understand and interact with said human; The system and a system AI agent designed to understand and interact with the system; Equipped with an architecture in which the human AI agent and the system AI agent are configured to communicate with each other, the human AI agent to learn about the system, and the system AI agent to learn about the human, to optimize interaction between the human and the system.
2. 2. The architecture of claim 1, wherein the human AI agent and the system AI agent are configured such that the human AI agent learns about the system and the system AI agent learns about the human during a configuration process prior to deploying the architecture.
3. 3. The architecture of claim 2, wherein the human AI agent and the system AI agent are configured such that, during operation of the architecture, the human AI agent learns about the system and the system AI agent learns about the human.
4. 4. The architecture of claim 3, wherein the human AI agent and the system AI agent iterate uncertainties during interaction between the human and the system to optimize the architecture for different environments and scenarios while the architecture is operational.
5. 10. The architecture of claim 1, wherein the human AI agent and the system AI agent exchange and analyze information such that together they identify capabilities and limitations of the human and the system and optimize the interaction between the human and the system to achieve a desired outcome.
6. 10. The architecture of claim 1, wherein 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.
7. 1. A method of providing an architecture, comprising: Designing a human artificial intelligence (AI) agent that understands and interacts with humans; Designing a system AI agent that understands and interacts with the system; providing communication between the human AI agent and the system AI agent such that the human AI agent learns about the system and the system AI agent learns about the human and optimizes interaction between the human and the system; A method comprising:
8. 8. The method of claim 7, wherein providing communication between the human AI agent and the system AI agent comprises having the human AI agent learn about the system and having the system AI agent learn about the human during a configuration process prior to deploying the architecture.
9. 10. The method of claim 8, wherein providing communication between the human AI agent and the system AI agent causes the human AI agent to learn about the system and the system AI agent to learn about the human during operation of the architecture.
10. 10. The method of claim 9, wherein providing communication between the human AI agent and the system AI agent causes the human AI agent and the system AI agent to iterate uncertainties during interaction between the human and the system so as to optimize the architecture for different environments and scenarios while the architecture is operational.
11. 8. The method of claim 7, wherein providing communication between the human AI agent and the System AI agent causes the human AI agent and the System AI agent to exchange and analyze information such that together they can identify capabilities and limitations of the human and the system and optimize interaction between the human and the system to achieve a desired outcome.
12. 8. The method of claim 7, wherein 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.
13. The architecture is means for providing a human artificial intelligence (AI) agent that understands and interacts with a human; means for providing a system AI agent for understanding and interacting with the system; means for providing communication between the human AI agent and the system AI agent such that the human AI agent learns about the system and the system AI agent learns about the human and optimizes interaction between the human and the system; An architecture that includes:
14. 14. The architecture of claim 13, wherein the means for providing communication between the human AI agent and the system AI agent allows the human AI agent to learn about the system and the system AI agent to learn about the human during a configuration process prior to deploying the architecture.
15. 15. The architecture of claim 14, wherein the means for providing communication between the human AI agent and the system AI agent causes the human AI agent to learn about the system and the system AI agent to learn about the human during operation of the architecture.
16. 15. The architecture of claim 14, wherein the means for providing communication between the human AI agent and the system AI agent allows the human AI agent and the system AI agent to iterate uncertainties during interaction between the human and the system so as to optimize the architecture for different environments and scenarios while the architecture is operating.
17. 15. The architecture of claim 14, wherein the means for providing communication between the human AI agent and the System AI agent causes the human AI agent and the System AI agent to exchange and analyze information such that together they can identify capabilities and limitations of the human and the system and optimize interactions between the human and the system to achieve desired outcomes.
18. 14. The architecture of claim 13, wherein 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.