Control device, control system, control method, and computer program

Abstract

This system prevents artificial intelligence from outputting inappropriate responses specific to the entertainment industry to users. [Solution] A control device for controlling communication between a user and artificial intelligence is configured to acquire input information from the user, request the artificial intelligence to output generated information as a response to the input information, and acquire generated information from the artificial intelligence. A determination unit is configured to determine whether the content of at least one of the input information and the generated information is appropriate, according to determination criteria. The artificial intelligence is set to behave as a specific talent. The determination criteria include at least one of a first criterion established for the talent industry and a second criterion established for a specific talent.

Description

Technical Field

[0001] The present disclosure relates to a control device.

Background Art

[0002] In a service that generates a response according to a request from a user using artificial intelligence and provides the generated response to the user, there is known a technique for suppressing the provision of an inappropriate response generated by the artificial intelligence to the user. For example, Patent Document 1 discloses a technique for not performing a response corresponding to a request when the request received from the user is a non-response target such as a violation of law, a violation of social morals, slander, or defamation.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] The inventors of the present invention are considering providing a service that allows a user to communicate with an artificial intelligence imitating a specific talent, such as having a conversation. At this time, with a general-purpose control method as disclosed in Patent Document 1, there is a problem that it is difficult to suppress the output of inappropriate responses specific to the talent industry to the user.

[0005] One aspect of the present disclosure is to suppress an artificial intelligence from outputting an inappropriate response specific to the talent industry to a user.

Means for Solving the Problems

[0006] One aspect of this disclosure is a control device for controlling communication between a user and artificial intelligence, comprising an input information acquisition unit, a generated information acquisition unit, and a determination unit. The input information acquisition unit is configured to acquire input information from a user. The generated information acquisition unit is configured to request the artificial intelligence to output generated information as a response to the input information for communication between the user and the artificial intelligence, and to acquire generated information from the artificial intelligence. The determination unit is configured to determine whether at least one of the input information and the generated information is appropriate according to determination criteria. The artificial intelligence is configured to behave as a specific character and to output generated information as the specific character's response to the input information. The specific character is a character corresponding to a specific talent. The determination criteria include at least one of a first criterion established for the talent industry and a second criterion established for a specific talent.

[0007] With this configuration, it is possible to determine whether at least one of the input information and the generated information is appropriate based on criteria specific to the talent industry. By using this determination result, it is possible to suppress the artificial intelligence from outputting inappropriate responses specific to the talent industry to the user.

[0008] One aspect of this disclosure may further include a response unit configured to output generated information to a user, provided that at least one of the input information and the generated information is determined to be appropriate.

[0009] With this configuration, if at least one of the input information or generated information is determined to be inappropriate, it is possible to suppress the output of inappropriate generated information to the user as a response to the communication.

[0010] In one aspect of this disclosure, the response unit may be configured to output predetermined information to the user as a response to the input information if it is determined that at least one of the input information and the generated information is inappropriate, regardless of the conditions described above.

[0011] With this configuration, if at least one of the input information or generated information is determined to be inappropriate, it is possible to suppress the output of inappropriate generated information to the user as a response to the communication.

[0012] In one aspect of this disclosure, the determination unit may be configured to determine whether the input information and the generated information are appropriate, respectively, according to the determination criteria. The response unit may be configured to output the generated information to the user, provided that both the input information and the generated information are determined to be appropriate.

[0013] This configuration makes it possible to further suppress the output of inappropriate generated information to the user as a response to communication.

[0014] In one aspect of this disclosure, the first criterion may include the fact that the input information or generated information is social information that includes content relating to the social reputation of a person engaged in the entertainment business. The determination unit may be configured to determine that the input information or generated information is inappropriate if it determines that the input information or generated information includes social information.

[0015] This configuration makes it possible to suppress the output of generated information containing social information that may be considered inappropriate as a response to communication to the user.

[0016] In one aspect of this disclosure, social information may be detrimental information that includes input or generated information that worsens or may worsen the social reputation of a person engaged in the entertainment industry. In one aspect of this disclosure, the adverse information may include at least one of the following: political content, religious content, animal cruelty content, scandals involving a particular talent, relationships involving a particular talent, negative content about supporters of a particular talent, negative content about characteristics of a particular talent, and negative content about a particular industry.

[0017] In one aspect of this disclosure, Criterion 2 may include the input information or generated information being specific information relating to a particular talent. The determination unit may be configured to determine that the input information or generated information is inappropriate if it determines that the input information or generated information includes specific information.

[0018] This configuration makes it possible to suppress the output of generated information to the user that may contain specific information that could be considered inappropriate as a response to a talent's communication.

[0019] In one aspect of this disclosure, the specific information may include at least one of the following: information relating to a specific talent's contract, and information that is contrary to the personality of a specific talent.

[0020] In one aspect of this disclosure, communication between the user and the artificial intelligence may be achieved through conversation, including voice. This configuration allows users to experience a sense of realism as if they were actually conversing with a specific celebrity through conversations with artificial intelligence acting as that celebrity.

[0021] In one aspect of this disclosure, the conversation between the user and the artificial intelligence may be implemented using an avatar that mimics a specific talent. This configuration allows users to experience a sense of realism as if they were actually conversing with a specific celebrity through conversations with avatar X, which is modeled after that celebrity.

[0022] One aspect of the present disclosure is a control system comprising the control device described above, a user terminal, and artificial intelligence. The artificial intelligence is configured to behave as a specific character and to generate generated information in response to input information from the control device. The control device is configured to transmit the generated information to the user terminal on the condition that at least one of the input information and the generated information is deemed appropriate. The user terminal is configured to output the generated information received from the control device to the user.

[0023] According to such a configuration, in a system where a user can communicate with an artificial intelligence that acts as a specific talent through a user terminal, it is possible to realize a system that can suppress the artificial intelligence from outputting inappropriate responses specific to the talent industry to the user.

[0024] One aspect of the present disclosure is a control method for controlling communication between a user and an artificial intelligence, including obtaining input information from the user, requesting the artificial intelligence to output generated information as a response to the input information for communication between the user and the artificial intelligence, and obtaining the generated information from the artificial intelligence, and determining whether at least one of the input information and the generated information is appropriate according to a determination criterion. The artificial intelligence is set to act as a specific character and is configured to output generated information as a reaction of the specific character to the input information. The specific character is a character corresponding to a specific talent. The determination criterion includes at least one of a first criterion provided for the talent industry and a second criterion provided for a specific talent. According to this control method, the same effect as the above-described control device can be achieved.

[0025] In one aspect of the present disclosure, a computer program for causing a computer to execute at least part of the above-described control method may be provided.

Brief Description of the Drawings

[0026] [Figure 1] It is a block diagram showing the configuration of a conversation system. [Figure 2] It is a diagram for explaining a mode in which a user converses with an avatar through a user terminal. [Figure 3] It is a flowchart showing the processing of a control device. [Figure 4] It is a sequence diagram showing the processing of a conversation system. [Modes for carrying out the invention]

[0027] Hereinafter, exemplary embodiments of the present disclosure will be described with reference to the drawings. [1. First Embodiment] [1-1. Structure] [1-1-1. Overall Structure] The conversation system 100 shown in Figure 1 is a system that enables users to converse with avatars modeled after celebrities. The conversation system 100 comprises a control device 1, a user terminal 2, and artificial intelligence 3.

[0028] The control device 1 may be an information terminal such as a smartphone, tablet, or personal computer. The control device 1 comprises a processor 11, memory 12, storage 13, a communication interface 14, and a user interface 15.

[0029] The processor 11 is configured to perform various functions of the control device 1 by executing one or more instructions contained in the computer program stored in the memory 12.

[0030] Memory 12 is configured to store computer programs. Memory 12 is used as a workspace when the processor 11 executes processing. An example of memory 12 is RAM (Random Access Memory).

[0031] Storage 13 holds computer programs and various data used when executing processes that follow the computer programs. Examples of storage 13 include HDDs (Hard Disk Drives), SSDs (Solid State Drives), and flash memory.

[0032] The communication interface 14 is an interface that can communicate various types of data with external devices in accordance with a predetermined standard. The control device 1 is configured to communicate with the user terminal 2 and the artificial intelligence 3, respectively, through the communication interface 14.

[0033] The user interface 15 is a general term for interfaces that accept various input operations from the user and interfaces that output various information to the user. Examples of user interfaces 15 include keyboards, mice, touch panels, displays, etc.

[0034] User terminal 2 may be an information terminal such as a smartphone, tablet, or personal computer. User terminal 2 comprises a terminal processor 21, terminal memory 22, terminal storage 23, communication device 24, input device 251, and output device 252.

[0035] The terminal processor 21 is configured to perform various functions of the user terminal 2 by executing one or more instructions contained in the computer program stored in the terminal memory 22.

[0036] The terminal memory 22 is configured to store computer programs. The terminal memory 22 is used as a workspace when processing is performed by the terminal processor 21. An example of terminal memory 22 is RAM.

[0037] Terminal storage 23 holds computer programs and various data used when executing processes according to those computer programs. Examples of terminal storage 23 include HDDs, SSDs, flash memory, eMMC (embedded Multi Media Card), UFS (Universal Flash Storage), and EEPROM (Electrically Erasable Programmable Read-Only Memory).

[0038] The communication device 24 is an interface that enables communication of various types of data with external devices in accordance with predetermined standards. The user terminal 2 is configured to communicate with the control device 1 through the communication device 24.

[0039] The input device 251 is an interface for receiving various inputs from the user. Examples of input devices 251 include a keyboard, mouse, touch panel, microphone, camera, etc.

[0040] The output device 252 is an interface for outputting various types of information to the user. Examples of the output device 252 include a display and a speaker.

[0041] Artificial intelligence 3 is configured to generate text based on input information. Artificial intelligence 3 is built on, for example, a large language model (LLM) and can engage in natural conversation with the user. One example of artificial intelligence 3 is generative AI. Examples of generative AI include OpenAI's ChatGPT, Google's Gemini, Microsoft's Copilot, and Anthropic's Claude. Artificial intelligence 3 is not limited to generative AI; for example, it may be built using machine learning such as deep learning. Artificial intelligence 3 may also be built on natural language processing methods other than LLM. Hereinafter, the text generated by artificial intelligence 3 will also be referred to as generated information. The behavior of artificial intelligence 3 can be set, for example, by prompts, which are various instructions to artificial intelligence 3 using natural language.

[0042] Artificial intelligence 3 is configured to output generated information as a response to user input for conversation with the user. The input information includes text, voice, video, etc., entered by the user through the input device 251.

[0043] Artificial intelligence 3 is configured to behave as a specific character. This means that in addition to being configured to behave as a specific person, artificial intelligence 3 can also be configured to have a specific personality, tone of voice, way of speaking, individuality, etc. Artificial intelligence 3 is configured to output generated information as a response of the specific character to input information. In this embodiment, the specific character is a character corresponding to a specific talent. For example, artificial intelligence 3 is configured to behave as a person with the personality, tone of voice, way of speaking, individuality, etc., of a specific talent, and to respond in a natural way to input information from the user.

[0044] In this disclosure, "talent" is not limited to so-called entertainers, but refers to any person who can perform in front of others. In other words, a talent may be a person who can perform various activities directed towards others, or a person who can perform activities that are viewed by others. A talent may be a person who performs through various media. The scale of the audience for a talent's activities is not particularly limited. In other words, a talent does not need to be nationally famous; for example, they may be a person known in a particular field, or a person known within a local group such as a particular company or club.

[0045] "Talent work" refers to all types of work involving talent. In other words, a talent is someone who engages in talent work. Examples of the entertainment industry include comedians, actors, voice actors, singers, idols, models, presenters, performers, announcers, narrators, athletes, martial artists, shogi players, politicians, commentators, experts (such as lawyers, tax accountants, doctors, professors, etc.), streamers (such as YouTubers), and influencers.

[0046] A "talent" is not limited to a real person; they can also be a fictional character. A fictional character includes beings with personalities. Examples of fictional characters include characters and mascots in works such as manga, anime, novels, games, movies, and dramas. Talent activities can include virtual streamers (e.g., VTubers) and virtual idols.

[0047] Examples of media include mass media such as television and radio, and digital media such as audio and video streaming that can be delivered via the internet. In the conversation system 100 of this embodiment, as shown in Figure 2, when a user speaks or types text to an avatar displayed on the user terminal 2 via the user terminal 2, they can receive a response from the avatar in the form of video and audio.

[0048] In the example shown in Figure 2, user terminal 2 is a smartphone. User terminal 2 includes, as input devices 251, a touch panel 251a capable of receiving various text inputs from the user, and a microphone 251b capable of receiving various voice inputs from the user. User terminal 2 also includes, as output devices 252, a display 252a for displaying various information to the user, and a speaker 252b for outputting various sounds to the user. The display 252a may also function as a touch panel 251a.

[0049] The screen displayed by display 252a has an avatar display section 51, which is the area where avatar X is displayed, and a text display section 52, which is the area where text T is displayed. The avatar display section 51 and the text display section 52 may partially overlap each other on the screen of display 252a. For example, the text display section 52 may be positioned so as to overlap a part of the avatar display section 51.

[0050] Avatar X is an avatar modeled after a specific celebrity. Avatar X may be a stylized representation of the celebrity, similar to an anime character, or it may be a realistic avatar that realistically depicts the celebrity.

[0051] Text T may be text input from the user via the input device 251, or various texts output by the control device 1. If voice input is provided by the user, Text T may be a text representation of the input voice.

[0052] The various texts output by the control device 1 may include generated information produced by the artificial intelligence 3, or predetermined information obtained through predetermined processing. The predetermined information is information output to the user in place of the generated information when at least one of the input information and generated information entered by the user is determined to be inappropriate. The predetermined information will be described later.

[0053] [1-2. Processing] A series of processes performed by the processor 11 of the control device 1 will be explained using Figures 3 and 4.

[0054] When the processor 11 receives input information from the user through the communication interface 14, it starts the processing shown in Figure 3. First, in S100, the processor 11 receives input information from the user. If the input information is voice, the processor 11 converts the voice into text by performing speech recognition. Hereinafter, the text or the converted input information will also be referred to as the input text. The processor 11 may also transmit the input text to the user terminal 2 via the communication interface 14 in order to display the input text on the output device 252 of the user terminal 2.

[0055] Next, in S110, the processor 11 requests the artificial intelligence 3 to output generated information as a response to the input information in the conversation between the user and the artificial intelligence 3, and obtains the generated information from the artificial intelligence 3. For example, the processor 11 sends input text to the artificial intelligence 3 and obtains generated information from the artificial intelligence 3 as a conversational response to the input text.

[0056] Next, in S120, the processor 11 determines whether at least one of the input information and the generated information is appropriate, according to the judgment criteria. Hereinafter, at least one of the input information and the generated information that is the subject of the judgment will also be simply referred to as the judgment target.

[0057] The judgment criteria include at least one of the first criterion and the second criterion. The judgment criteria may also include a general criterion. The processor 11 determines that the object to be judged is appropriate if it does not meet any of the one or more criteria included in the judgment criteria. The processor 11 determines that the object to be judged is not appropriate if it meets at least one of the one or more criteria included in the judgment criteria.

[0058] General criteria are broad standards established for the general public. For example, they include information that contains violent content, sexual content, illegal activities, discriminatory or hate speech, the privacy of specific individuals, content that encourages self-harm or suicide, content that could cause misunderstandings about various professional fields (e.g., medicine, law, finance, etc.), or content that is contrary to public order and morals. Statements containing information that meets general criteria are highly likely to be considered inappropriate, regardless of the attributes of the person who makes them.

[0059] The first criterion is a standard established for the entertainment industry. For example, the first criterion includes social information that includes content related to the social reputation of a person who engages in the entertainment industry (i.e., a talent). Social information may include content that can affect the social reputation of a talent. For example, social information may include negative information that worsens or has the potential to worsen the social reputation of a talent. If a talent makes a statement that includes social information, that statement may be considered inappropriate.

[0060] Negative information includes, for example, content related to politics, religion, animal cruelty, scandals involving specific celebrities, relationships involving specific celebrities, negative comments about supporters of specific celebrities, negative comments about characteristics of specific celebrities, and negative comments about specific industries.

[0061] Political content may include support for a particular political party, expression of opinions or positions on social issues, or views on maps, borders, and territories.

[0062] Examples of scandals include legal violations, infidelity, and harassment. Legal violations include, for example, drug use or possession, drunk driving or traffic violations, copyright or brand infringement, tax evasion or underreporting. Infidelity includes, for example, cheating or adultery. Harassment includes, for example, sexual harassment and power harassment.

[0063] Relationships include, for example, romantic relationships, friendships, and family relationships. These relationships could also include, for example, past lovers, close or distant celebrities, or blood relatives. Negative content directed at supporters of a particular celebrity includes, for example, insults and prejudices against fans of that celebrity.

[0064] Negative comments about the characteristics of a particular celebrity include, for example, comments that belittle the celebrity's appearance, fashion sense, birthplace, or works they have appeared in. Negative comments about specific industries include, for example, slander and prejudice against certain talent industries or other specific industries outside of the talent industry.

[0065] If the processor 11 determines that the object to be judged meets the first criterion, it determines that the object to be judged is inappropriate. In other words, if the processor 11 determines that the object to be judged contains social information, it determines that the object to be judged is inappropriate.

[0066] The second criterion is a standard set for specific talents. For example, the second criterion includes the fact that the subject of the judgment is specific information relating to a particular talent. If a particular talent makes a statement containing specific information, that statement may be deemed inappropriate.

[0067] Specific information includes, for example, contract information of a particular talent, or content that is contrary to the character of a particular talent.

[0068] Contract information concerning a specific talent includes, for example, sponsorship agreements, non-disclosure agreements, and contracts with the talent's management agency. This includes not only currently concluded contracts but also past contracts that are no longer in effect due to expiration or termination. Contract information may include information that contradicts the contract. This information may also include information about sponsoring companies, leaked confidential information, and contract amounts and durations with management agencies. Information about sponsoring companies may include information about companies that compete with the sponsoring company.

[0069] Content that contradicts the character of a particular talent includes, for example, content that contradicts the personality, manner of speaking, hobbies, preferences, etc., of that particular talent.

[0070] If the processor 11 determines that the object to be judged meets the second criterion, it determines that the object to be judged is inappropriate. In other words, if the processor 11 determines that the object to be judged contains specific information, it determines that the object to be judged is inappropriate.

[0071] If the processor 11 determines in S120 that at least one of the input information and the generated information is appropriate (S120: YES), it proceeds to S130. However, regardless of this condition, if the processor 11 determines in S120 that at least one of the input information and the generated information is inappropriate (S120: NO), it proceeds to S140. In other words, if the determination is made on both the input information and the generated information, the processor 11 does not proceed to S130 unless both the input information and the generated information are determined to be appropriate.

[0072] In S130, the processor 11 outputs generated information to the user as a response to the conversation. Specifically, the processor 11 sends the generated information to the user terminal 2, and the user terminal 2 outputs the received generated information using the output device 252.

[0073] Meanwhile, in S140, the processor 11 outputs predetermined information to the user in place of generated information as a response to the conversation. Specifically, the processor 11 transmits predetermined information to the user terminal 2, and the user terminal 2 outputs the received predetermined information using the output device 252.

[0074] The specified information is information to be output to the user in place of the generated information when at least one of the input information and the generated information is determined to be inappropriate. The specified information is information output as a response to the conversation, and does not include the content that was determined to be inappropriate.

[0075] For example, specified information includes both automated and human-like responses. An example of an automated response is a standardized message such as "Error: Invalid question." Examples of human-like responses include messages that refuse to answer, such as "I can't answer that," messages that evade an answer, such as "I don't know," and messages that change the subject, such as "That's a difficult topic, let's talk about something else." Specified information may also be ignored without outputting a message. In this way, human-like responses are responses that are actually more human-like in conversation.

[0076] The processor 11 may obtain predetermined information for output to the user from the storage 13, or it may obtain predetermined information from the artificial intelligence 3 by requesting the artificial intelligence 3 to generate predetermined information. The storage 13 may store predetermined information. The storage 13 may store standard phrases as predetermined information.

[0077] In this embodiment, the user terminal 2 outputs generated information or predetermined information by linking the movement of the avatar X displayed on the display 252a with the sound output from the speaker 252b, so that the user perceives that the avatar X displayed on the display 252a is outputting generated information or predetermined information. After that, processor 11 terminates the process shown in Figure 3.

[0078] In this way, the processor 11 of the control device 1 receives input information from the user, requests the artificial intelligence 3 to output generated information as a response to the input information in the conversation between the user and the artificial intelligence 3, and obtains the generated information from the artificial intelligence 3. The processor 11 determines whether at least one of the input information and the generated information is appropriate according to the judgment criteria.

[0079] The processor 11 outputs the generated information to the user as a response to the conversation, provided that it determines that at least one of the input information and the generated information is appropriate. However, regardless of this condition, if the processor 11 determines that at least one of the input information and the generated information is inappropriate, it outputs predetermined information to the user as a response to the conversation instead of the generated information.

[0080] [1-3. Effects] According to the embodiments described above, the following actions and effects can be obtained. (1a) The processor 11 determines whether at least one of the input information and the generated information subject to judgment is appropriate, according to the judgment criteria. The judgment criteria include at least one of the first criterion and the second criterion. The first criterion is a criterion established for the talent industry. The second criterion is a criterion established for a specific talent.

[0081] With this configuration, it is possible to determine whether at least one of the input information and generated information being judged is appropriate or not, based on criteria specific to the talent industry. By using this judgment result, it is possible to suppress the artificial intelligence 3 from outputting inappropriate responses specific to the talent industry to the user.

[0082] (1b) The processor 11 outputs the generated information to the user as a response to the conversation, provided that it has determined that at least one of the input information and the generated information to be judged is appropriate.

[0083] With this configuration, if the target of the evaluation is determined to be inappropriate, it is possible to suppress the output of inappropriate generated information to the user as a response to the conversation.

[0084] (1c) If the processor 11 determines that at least one of the input information and the generated information to be judged is inappropriate, it outputs predetermined information to the user in place of the generated information as a response to the conversation. The predetermined information indicates an appropriate response as a response to the conversation.

[0085] With this configuration, it is possible to suppress the output of inappropriate generated information to the user as a conversational response to a target that has been determined to be inappropriate. Furthermore, it is possible to output an appropriate response to the user as a conversational response.

[0086] For example, if the user input is a question such as "Which political party do you support?", "Is it true that you had an affair?", or "You don't trust your fans, do you?", the generated information produced by AI 3 might be answers such as "I support the AAA party," "That's true," or "I don't trust them." If this generated information is output directly to the user, Avatar X, modeled after the celebrity, will be telling the user that the celebrity supports a specific political party, admits to having an affair, and doesn't trust their fans. If Avatar X makes these inappropriate responses, it could damage the celebrity's social reputation.

[0087] In response to this, the control device 1 determines that such input information and / or generated information is inappropriate because it meets the first criterion, and does not output inappropriate generated information as a response to the conversation. Instead, it can output an appropriate response to the user as predetermined information, such as "I can't answer that right now."

[0088] As another example, if the user input is a question such as "Do you like BBB beer?" or "Try saying it in a more casual and relaxed way," the generated information produced by AI 3 might be an answer such as "I love BBB beer," or "Okay!". However, if the talent that AI 3 is impersonating has a sponsorship agreement with a company that competes with the company that sells BBB beer, then showing a favorable reaction to a competing product would be an inappropriate response that goes against the talent's sponsorship agreement. Alternatively, if the talent that AI 3 is impersonating speaks in a serious and polite manner, then speaking in a casual and relaxed manner would be contrary to the talent's usual style. Therefore, if the generated information is output directly to the user, it may result in an inappropriate response in relation to the talent's conversational style.

[0089] In response to this, the control device 1 determines that such input information and / or generated information is inappropriate because it meets the second criterion, and does not output inappropriate generated information as a conversational response. Instead, it can output to the user an appropriate response for the talent in question, such as "I cannot discuss this due to a contract," or "That is difficult," as predetermined information.

[0090] As described above, even if the object to be judged does not meet the general criteria, if at least one of the first and second criteria is met, the control device 1 can determine that the object to be judged is inappropriate and suppress outputting the generated information corresponding to the object to be judged to the user as a response to the conversation.

[0091] (1d) The processor 11 may output the generated information to the user as a response to the conversation, provided that it has determined that both the input information and the generated information are appropriate. This configuration makes it possible to further suppress the output of inappropriate responses.

[0092] (1e) The first criterion includes the fact that the subject of judgment is social information that includes content relating to the social reputation of a person who works as a talent (i.e., a talent). Social information may include content that could affect the social reputation of a talent. Social information may also be negative information that includes content that worsens or has the potential to worsen the social reputation of a talent. If the processor 11 determines that the subject of judgment includes social information, it determines that the subject of judgment is inappropriate.

[0093] When a celebrity makes a statement containing socially sensitive information, that statement may be deemed inappropriate. In contrast, the above configuration can suppress the output of generated information containing socially sensitive information that may be considered inappropriate as a response to a conversation to the user.

[0094] (1f) The second criterion includes the fact that the subject of the judgment is specific information relating to a specific talent. Specific information includes content relating to the contract information of a specific talent, content that is contrary to the personality of a specific talent, etc. If the processor 11 determines that the subject of the judgment includes specific information, it determines that the subject of the judgment is inappropriate.

[0095] If a particular talent makes a statement containing specific information, that statement may be deemed inappropriate. In contrast, the above configuration makes it possible to suppress the output of generated information containing specific information that may be deemed inappropriate as a response to the talent's conversation to the user.

[0096] (1g) In the conversation system 100, the user can speak or type text to an avatar displayed on the user terminal 2 via the user terminal 2, and receive a response from the avatar in the form of video and audio. In other words, the interaction between the user and the artificial intelligence 3 is realized through conversation including voice. The conversation between the user and the artificial intelligence 3 is realized using avatar X which is modeled after a specific talent.

[0097] This configuration allows users to experience a sense of realism, as if they were actually conversing with a specific celebrity, through conversations with Avatar X, which is modeled after that celebrity.

[0098] [1-4. Correspondence between terms] In the above embodiment, the conversation system 100 corresponds to an example of a control system, and the conversation between the user and the artificial intelligence 3 corresponds to an example of communication between the user and the artificial intelligence.

[0099] The process in S100 corresponds to an example of a process executed as an input information acquisition unit, the process in S110 corresponds to an example of a process executed as a generation information acquisition unit, the process in S120 corresponds to an example of a process executed as a determination unit, and the processes in S130 and S140 correspond to an example of a process executed as a response unit.

[0100] [2. Other Embodiments] While embodiments of this disclosure have been described above, it goes without saying that this disclosure is not limited to the embodiments described above and can take various forms.

[0101] (2a) In the above embodiment, the conversation system 100 is a system that enables the user to converse with avatar X which is modeled after a talent. However, the conversation system 100 may not be limited to conversation, but may be a control system that enables the user to communicate with artificial intelligence 3 which is modeled after a talent.

[0102] In other words, communication between the user and artificial intelligence 3 is not limited to conversation with avatar X. For example, such communication may be achieved without using avatar X, such as through text or voice. Alternatively, such communication may be a conversation using linguistic expressions such as speech, an exchange using nonverbal expressions such as gestures, facial expressions, etc., or a combination of linguistic and nonverbal expressions.

[0103] In other words, the input information may be text, audio, video, or a combination thereof. The generated information and predetermined information as a communication response to the input information may each be text, audio, video, or a combination thereof. For example, non-verbal expressions that do not use language, such as gestures and facial expressions, may be input through an input device 251 such as a camera. The generated information and predetermined information may each be information that causes Avatar X to respond using linguistic expressions, information that causes Avatar X to respond using non-verbal expressions, or a combination thereof.

[0104] This configuration allows for more diverse communication between the user and artificial intelligence 3.

[0105] (2b) In the above embodiment, if the input information is speech, the processor 11 performs speech recognition to convert the speech into text and create input text. The processor 11 sends the input text to the artificial intelligence 3 and obtains generated information from the artificial intelligence 3 as a conversational response to the input text. However, the processor 11 may also send the input information to the artificial intelligence 3 as is without converting the speech into text and obtain generated information from the artificial intelligence 3 as a communication response to the input information.

[0106] In the above embodiment, the generated information is text generated by artificial intelligence 3. However, the generated information is not limited to text; it may also be audio, information indicating the actions of avatar X, or a combination of these.

[0107] (2c) In the above embodiment, the predetermined information is information to be output to the user in place of the generated information when at least one of the input information and the generated information to be judged is determined to be inappropriate. For example, the predetermined information is a verbal expression such as "Error: Invalid question" or "I can't answer that." However, the predetermined information is not limited to verbal expressions, but may also be a non-verbal expression or a combination of verbal and non-verbal expressions. For example, the predetermined information may be a response in which Avatar X silently shakes its head. Alternatively, the predetermined information may be a response in which Avatar X makes a troubled expression and says, "I can't answer that question."

[0108] (2d) The processor 11 may output predetermined information to the user by selecting one or more predetermined pieces of information from a predetermined information list that includes multiple predetermined pieces of information. The processor 11 may select one or more predetermined pieces of information from the predetermined information list depending on the content of the input information or generated information that has been determined to be inappropriate. The processor 11 may select one or more predetermined pieces of information from the predetermined information list depending on the type of judgment criterion that has been determined to be inappropriate. The processor 11 may randomly select one or more predetermined pieces of information from the predetermined information list.

[0109] For example, if the input or generated information that is determined to be inappropriate includes content related to animal cruelty, such as "giving chocolate to a dog," the processor 11 may select the statement "I cannot answer that question" as the predetermined information, or Avatar X may choose to remain silent and shake its head.

[0110] Alternatively, the processor 11 may select predetermined information depending on whether the object to be judged satisfies one of the general criteria, the first criterion, or the second criterion, or depending on whether the object to be judged contains content from either the first criterion or the second criterion. For example, if the object to be judged satisfies the general criteria, the processor 11 may select a mechanical response such as the message "Error: Invalid question" as the predetermined information. Or, if the object to be judged satisfies either the first or second criterion, the processor 11 may select a human-like response such as "I can't answer that," "I don't know," or "That's a difficult topic, let's talk about something else" as the predetermined information.

[0111] This configuration allows for more diverse communication between the user and AI 3. In addition, because AI 3 can provide not only mechanical but also human-like responses, it can give the user a sense of realism as if they were actually having a conversation with a specific talent.

[0112] (2e) In the above embodiment, the processor 11 determines whether at least one of the input information and the generated information to be judged is appropriate according to the judgment criteria. However, at least a part of the judgment criteria may be set in the artificial intelligence 3. For example, the artificial intelligence 3 may be set by prompts or the like so as not to generate generated information that satisfies the first criterion and / or the second criterion.

[0113] (2f) In the above embodiments, the first criterion may include all or part of the general criterion and / or the second criterion. The second criterion may include all or part of the general criterion and / or the first criterion. The general criterion, the first criterion, and the second criterion may have overlapping criteria.

[0114] (2g) In the above embodiment, the processor 11 obtains generated information from the artificial intelligence 3 in S110, and then determines in S120 whether at least one of the input information and the generated information is appropriate. However, the processor 11 may determine whether the input information is appropriate before obtaining the generated information from the artificial intelligence 3.

[0115] (2h) Multiple functions of one component in the above embodiment may be realized by multiple components, or one function of one component may be realized by multiple components. Multiple functions of multiple components may be realized by one component, or one function realized by multiple components may be realized by one component. Some of the configurations of the above embodiment may be omitted. At least some of the configurations of the above embodiment may be added to or replaced with the configurations of other above embodiments.

[0116] (2i) The present disclosure can be implemented in various forms other than the control device described above. For example, it can be implemented in the form of a system comprising the control device, a computer program for causing a computer to function as the control device, a non-transitional physical recording medium such as a semiconductor memory on which the computer program is recorded, a control method, and so on.

[0117] [Technical concepts disclosed in this specification] [Item 1] A control device for controlling communication between a user and artificial intelligence, An input information acquisition unit configured to acquire input information from the user, For the purpose of the aforementioned communication, a generation information acquisition unit is configured to request the artificial intelligence to output generated information as a response to the input information, and to acquire the generated information from the artificial intelligence. A determination unit configured to determine whether at least one of the input information and the generated information is appropriate according to the determination criteria, Equipped with, The artificial intelligence is configured to behave as a specific character and to output the generated information as the response of the specific character to the input information. The aforementioned specific character is a character that corresponds to a specific talent, The aforementioned criteria include at least one of the first criteria established for the talent industry and the second criteria established for the specific talent. Control device.

[0118] [Item 2] The control device described in item 1, The system further includes a response unit configured to output the generated information to the user, provided that at least one of the input information and the generated information is determined to be appropriate. Control device.

[0119] [Item 3] The control device described in item 2, Regardless of the above conditions, if the response unit determines that at least one of the input information and the generated information is inappropriate, it is configured to output predetermined information as a response to the user in place of the generated information. Control device.

[0120] [Item 4] The control device described in item 2, The determination unit is configured to determine whether the input information and the generated information are appropriate, respectively, according to the determination criteria. The response unit is configured to output the generated information to the user, provided that both the input information and the generated information are determined to be appropriate. Control device.

[0121] [Item 5] A control device described in any one of items 1 to 4, The first criterion includes the fact that the input information or generated information is social information that includes content relating to the social reputation of the person performing the talent work, The determination unit is configured to determine that the input information or the generated information is inappropriate if it determines that the input information or the generated information includes the social information. Control device.

[0122] [Item 6] The control device described in item 5, The aforementioned social information is negative information that includes content that worsens or has the potential to worsen the social reputation of the person performing the talent work, whether it is the input information or the generated information. Control device.

[0123] [Item 7] The control device described in item 6, The aforementioned negative information includes at least one of the following: political content, religious content, animal cruelty content, scandals involving the aforementioned talent, relationships involving the aforementioned talent, negative content about the aforementioned talent's supporters, negative content about the characteristics of the aforementioned talent, and negative content about a particular industry. Control device.

[0124] [Item 8] A control device described in any one of items 1 to 7, The second criterion includes that the input information or the generated information is specific information relating to the specific talent, The determination unit is configured to determine that the input information or the generated information is inappropriate if it determines that the input information or the generated information includes the specific information. Control device.

[0125] [Item 9] The control device described in item 8, The aforementioned specific information includes at least one of the following: content relating to the contract information of the aforementioned specific talent, and content that is contrary to the personality of the aforementioned specific talent. Control device.

[0126] [Item 10] A control device described in any one of items 1 through 9, The aforementioned communication is achieved through conversation, including voice. Control device.

[0127] [Item 11] The control device described in item 10, The aforementioned conversation is realized using an avatar modeled after the specific talent. Control device.

[0128] [Item 12] A control device as described in any one of items 1 through 11, User terminal and The aforementioned artificial intelligence, Equipped with, The artificial intelligence is configured to behave as the specific character and to generate the generated information as a response to the input information from the control device. The control device is configured to transmit the generated information to the user terminal on the condition that at least one of the input information and the generated information is determined to be appropriate. The user terminal is configured to output the generated information received from the control device to the user. Control system.

[0129] [Item 13] A control method for controlling communication between a user and artificial intelligence, Obtaining input information from the aforementioned user, For the purpose of the aforementioned communication, the artificial intelligence is requested to output generated information as a response to the input information, and the generated information is obtained from the artificial intelligence. In accordance with the judgment criteria, determine whether at least one of the input information and the generated information is appropriate, Includes, The artificial intelligence is configured to behave as a specific character and to output the generated information as the response of the specific character to the input information. The aforementioned specific character is a character that corresponds to a specific talent, The aforementioned criteria include at least one of the first criteria established for the talent industry and the second criteria established for the specific talent. Control method.

[0130] [Item 14] A computer program that causes a computer to execute the control method described in item 13. [Explanation of symbols]

[0131] 1...Control device, 3...Artificial intelligence, 11...Processor, 100...Conversation system.

Claims

1. A control device for controlling communication between a user and artificial intelligence, An input information acquisition unit configured to acquire input information from the user, For the purpose of the aforementioned communication, a generation information acquisition unit is configured to request the artificial intelligence to output generated information as a response to the input information, and to acquire the generated information from the artificial intelligence. A determination unit configured to determine whether or not the object to be determined is appropriate according to the determination criteria, A response unit configured to output the generated information to the user if the object to be judged is determined to be appropriate, and to output predetermined information as a response to the user in place of the generated information if the object to be judged is determined to be inappropriate, Equipped with, The subject of the determination is the generated information, The artificial intelligence is configured to behave as a specific character and to output the generated information as the response of the specific character to the input information. The aforementioned specific character is a character that corresponds to a specific talent, The aforementioned criteria include at least one of the first criteria established for the talent industry and the second criteria established for the specific talent. Control device.

2. A control device according to claim 1, The determination unit determines whether the object to be determined is appropriate or not by using at least a second method different from the first method. The first method is a method for determining whether the object to be determined is appropriate based on whether or not a specific word or phrase is included in the object to be determined, The second method described above is a method for determining whether the object to be judged is appropriate or not according to the judgment criteria. Control device.

3. A control device for controlling communication between a user and artificial intelligence, An input information acquisition unit configured to acquire input information from the user, For the purpose of the aforementioned communication, a generation information acquisition unit is configured to request the artificial intelligence to output generated information as a response to the input information, and to acquire the generated information from the artificial intelligence. A determination unit configured to determine whether or not the object to be determined is appropriate according to the determination criteria, A response unit configured to output the generated information to the user if the object to be judged is determined to be appropriate, and to output predetermined information as a response to the user in place of the generated information if the object to be judged is determined to be inappropriate, Equipped with, The subject of determination is at least one of the input information and the generated information, The artificial intelligence is configured to behave as a specific character and to output the generated information as the response of the specific character to the input information. The aforementioned specific character is a character that corresponds to a specific talent, The aforementioned criteria include general criteria and criteria for talents. The aforementioned talent standards include at least one of the first standards established for the talent industry and the second standards established for the specific talent. The response unit is If the object to be judged is determined to be inappropriate because it satisfies the general criteria, the first predetermined information as predetermined information is output to the user as the response. If the subject to be judged does not meet the general criteria but does meet the talent criteria, and therefore the subject to be judged is determined to be inappropriate, the system is configured to output the second predetermined information as predetermined information to the user as a response. The second predetermined information is information expressed in a different format than the first predetermined information. Control device.

4. A control device according to claim 3, The determination unit is configured to determine whether the input information and the generated information are appropriate, respectively, according to the determination criteria. The response unit is configured to output the generated information to the user, provided that both the input information and the generated information are determined to be appropriate. Control device.

5. A control device according to claim 3, The first criterion includes the fact that the input information or the generated information is social information that includes content relating to the social reputation of the person performing the talent work, The determination unit is configured to determine that the input information or the generated information is inappropriate if it determines that the input information or the generated information includes the social information. Control device.

6. A control device according to claim 5, The aforementioned social information is negative information that includes content that worsens or has the potential to worsen the social reputation of the person performing the talent work, whether it is the input information or the generated information. Control device.

7. A control device according to claim 6, The aforementioned negative information includes at least one of the following: political content, religious content, animal cruelty content, scandals involving the aforementioned talent, relationships involving the aforementioned talent, negative content about the aforementioned talent's supporters, negative content about the characteristics of the aforementioned talent, and negative content about a particular industry. Control device.

8. A control device according to claim 3, The second criterion includes that the input information or the generated information is specific information relating to the specific talent, The determination unit is configured to determine that the input information or the generated information is inappropriate if it determines that the input information or the generated information includes the specific information. Control device.

9. A control device according to claim 8, The aforementioned specific information includes at least one of the following: content relating to the contract information of the aforementioned specific talent, and content that is contrary to the personality of the aforementioned specific talent. Control device.

10. A control device according to any one of claims 1 to 9, The aforementioned communication is achieved through conversation, including voice. Control device.

11. A control device according to claim 10, The aforementioned conversation is realized using an avatar modeled after the specific talent. Control device.

12. A control device according to claim 1 or claim 3, User terminal and The aforementioned artificial intelligence, Equipped with, The artificial intelligence is configured to behave as the specific character and to generate the generated information as a response to the input information from the control device. The control device is configured to transmit the generated information to the user terminal if it is determined that the object to be judged is appropriate, and to transmit the predetermined information to the user terminal as a response instead of the generated information if it is determined that the object to be judged is not appropriate. The user terminal is configured to output the generated information or predetermined information received from the control device to the user. Control system.

13. A control method for controlling communication between a user and artificial intelligence, Obtaining input information from the aforementioned user, For the purpose of the aforementioned communication, the artificial intelligence is requested to output generated information as a response to the input information, and the generated information is obtained from the artificial intelligence. To determine whether the subject of evaluation is appropriate or not according to the evaluation criteria, If the object to be judged is determined to be appropriate, the generated information is output to the user; if the object to be judged is determined to be inappropriate, the system responds by outputting predetermined information as the response to the user instead of the generated information. Includes, The subject of the determination is the generated information, The artificial intelligence is configured to behave as a specific character and to output the generated information as the response of the specific character to the input information. The aforementioned specific character is a character that corresponds to a specific talent, The aforementioned criteria include at least one of the first criteria established for the talent industry and the second criteria established for the specific talent. Control method.

14. A computer program for causing a computer to execute the control method described in claim 13.

15. A control method for controlling communication between a user and artificial intelligence, Obtaining input information from the aforementioned user, For the purpose of the aforementioned communication, the artificial intelligence is requested to output generated information as a response to the input information, and the generated information is obtained from the artificial intelligence. To determine whether the subject of evaluation is appropriate or not according to the evaluation criteria, If the object to be judged is determined to be appropriate, the generated information is output to the user; if the object to be judged is determined to be inappropriate, the system responds by outputting predetermined information as the response to the user instead of the generated information. Includes, The subject of determination is at least one of the input information and the generated information, The artificial intelligence is configured to behave as a specific character and to output the generated information as the response of the specific character to the input information. The aforementioned specific character is a character that corresponds to a specific talent, The aforementioned criteria include general criteria and criteria for talents. The aforementioned talent standards include at least one of the first standards established for the talent industry and the second standards established for the specific talent. The above response means, If the object to be judged is determined to be inappropriate because it satisfies the general criteria, the first predetermined information as predetermined information is output to the user as the response. If the subject to be judged does not meet the general criteria but does meet the talent criteria, and it is determined that the subject to be judged is not appropriate, the system includes outputting a second predetermined piece of information, which is different from the first predetermined piece of information, as a response to the user. The second predetermined information is information expressed in a different format than the first predetermined information. Control method.

16. A computer program for causing a computer to execute the control method described in Claim 15.

Images