Method

Abstract

To improve the technology used for adjusting the rights related to AI-generated content.SOLUTION: An information processing device 20 stores each of multiple pieces of content provided by multiple creators as material content that can be used for training a generative AI. Next, the information processing device 20 assigns a label to each of the multiple pieces of material content, the label including a content ID or creator ID. The information processing device 20 then trains a first classification AI using supervised learning with the multiple pieces of material content as training data, the first classification AI outputting a first probability that input content belongs to each class classified by the content ID or creator ID. After that, the information processing device 20 infers the first probability that AI-generated content belongs to each class using the first classification AI. The information processing device 20 then determines the reward to be granted to each of the multiple creators based on the inferred first probability.SELECTED DRAWING: Figure 4

Description

[Technical Field]

[0001] The present disclosure relates to a method. [Background technology]

[0002] Conventionally, techniques for adjusting rights relationships for AI-generated content have been known. For example, Patent Document 1 discloses determining the contribution rate of material data to AI-generated content and distributing financial benefits arising from the number of plays, sales, etc. of the AI-generated content to the rights holders of the material data according to the contribution rate. Regarding the determination of the contribution rate, it also discloses that, for example, when a new AI-generated song is generated using multiple songs (material data), if the degree of influence of each song (such as the length of the melody used in the AI-generated song) is known, the contribution rate is set according to the degree of influence. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 7414015 Summary of the Invention [Problem to be solved by the invention]

[0004] There was room for improvement in the technology used to handle rights for AI-generated content from a machine learning perspective.

[0005] The purpose of this disclosure, made in light of these circumstances, is to improve the technology used to adjust rights for AI-generated content. [Means for solving the problem]

[0006] According to one embodiment of the present disclosure, a method comprises: A method executed by an information processing device, Storing each of the plurality of pieces of content provided by the plurality of creators as material content that can be used for learning the generative AI; assigning a label including a content ID or a creator ID to each of the plurality of material contents; training a first classification AI that outputs a first probability that an input content belongs to each class classified by content ID or creator ID through supervised learning using the plurality of material contents as learning data; Inferring, by the first classification AI, a first probability that the AI-generated content generated by the user using the generation AI belongs to each of the classes; determining a reward to be granted to each of the plurality of creators based on the inferred first probability; Includes. [Effects of the Invention]

[0007] According to one embodiment of the present disclosure, techniques used to adjust rights relationships for AI-generated content are improved. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a block diagram illustrating a schematic configuration of a system according to an embodiment of the present disclosure. [Figure 2] FIG. 2 is a block diagram showing a schematic configuration of a terminal device. [Figure 3] FIG. 1 is a block diagram showing a schematic configuration of an information processing device. [Figure 4] 10 is a flowchart showing a first operation of the information processing device. [Figure 5] 10 is a flowchart showing a second operation of the information processing device. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, embodiments of the present disclosure will be described.

[0010] (Outline of the embodiment) An overview of a system 1 according to an embodiment of the present disclosure will be described with reference to Fig. 1. The system 1 includes a plurality of terminal devices 10 and an information processing device 20. Each of the terminal devices 10 and the information processing device 20 is communicably connected to a network 30 including, for example, the Internet and a mobile communication network.

[0011] The terminal device 10 is a computer such as a PC (Personal Computer), a smartphone, or a tablet terminal. In this embodiment, the terminal device 10 is used by a user who generates AI-generated content using a generation AI. Alternatively, the terminal device 10 is used by a creator who uploads content to the information processing device 20 as material content that can be used for training the generation AI. In this embodiment, the "content" refers to, for example, but is not limited to, images, and may be electronic data in any format, such as music, text, or a program. Furthermore, the "creator" refers to, for example, but is not limited to, the creator of the content, and will be described as referring to the copyright holder of the content.

[0012] The information processing device 20 is, for example, one server device or multiple computers that can communicate with each other. The information processing device 20 can communicate with each of the terminal devices 10 via a network 30. In this embodiment, the information processing device 20 functions as an AI platform. The AI ​​platform provides any services related to the generation AI, such as providing the generation AI to users, training the generation AI, and selling AI-generated content by the generation AI. In this embodiment, the AI ​​platform is implemented as a web application. In one example, a user uses the terminal device 10 to log in as a user to a web page provided by the information processing device 20. On the web page, the user selects desired material content provided by the AI ​​platform to train the generation AI. The user specifies desired prompts to cause the generation AI to generate AI-generated content. Training the generation AI and generating the AI-generated content are performed by the information processing device 20. The user can then download the AI-generated content to the terminal device 10 or sell it on the AI ​​platform. Meanwhile, a creator uses the terminal device 10 to log in as a creator to a web page provided by the information processing device 20. On the web page, the creator can select any content that he or she wishes to provide as material content from among the content stored in the terminal device 10 and upload it to the information processing device 20.

[0013] First, an overview of this embodiment will be described, and details will be provided later. The information processing device 20 stores each of multiple pieces of content provided by multiple creators as material content available for training of the generation AI. Next, the information processing device 20 assigns a label including a content ID or a creator ID to each of the multiple pieces of material content. Next, the information processing device 20 trains a first classification AI, which outputs a first probability that input content belongs to each class classified by the content ID or creator ID, through supervised learning using the multiple material contents as training data. Thereafter, the information processing device 20 infers, using the first classification AI, a first probability that AI-generated content generated by a user using the generation AI belongs to each class. Then, the information processing device 20 determines a reward to be granted to each of the multiple creators based on the inferred first probability.

[0014] Typically, a generation AI is trained to learn features from a large number of pieces of content used as training data. However, humans cannot intuitively recognize what features a generation AI has learned, making it difficult for humans to quantitatively and accurately determine the contribution of material data to AI-generated content. In contrast, according to this embodiment, a first classification AI is trained through supervised learning using multiple pieces of material content assigned a content ID or creator ID as training data. This first classification AI then infers a first probability that the AI-generated content belongs to each class classified by the content ID or creator ID, and determines the remuneration to be paid to each of multiple creators based on the inferred first probability. Therefore, for example, a portion of the profits generated by the sale of the AI-generated content can be distributed quantitatively and accurately to the creators of the material content as remuneration based on the contribution of the material content to the AI-generated content. Therefore, according to this embodiment, technology used to adjust rights relationships for AI-generated content is improved in that profits from AI-generated content can be distributed quantitatively and accurately to rights holders based on the contribution of the material content to the AI-generated content using machine learning.

[0015] Next, each component of the system 1 will be described in detail.

[0016] (Terminal Device Configuration) As shown in FIG. 2, the terminal device 10 includes a communication unit 11, an output unit 12, an input unit 13, a storage unit 14, and a control unit 15.

[0017] The communication unit 11 includes one or more communication interfaces connected to the network 40. The communication interfaces correspond to mobile communication standards such as, but not limited to, 4G (4th Generation) or 5G (5th Generation). In this embodiment, the terminal device 10 communicates with the information processing device 20 via the communication unit 11 and the network 30.

[0018] The output unit 12 includes one or more output devices that output information. Examples of the output devices include, but are not limited to, a display that outputs information as a video or a speaker that outputs information as a sound. Alternatively, the output unit 12 may include an interface for connecting an external output device.

[0019] The input unit 13 includes one or more input devices that detect an input operation by a user. Examples of the input devices include, but are not limited to, physical keys, capacitive keys, a mouse, a touch panel, a touch screen integrated with the display of the output unit 12, or a microphone. Alternatively, the input unit 13 may include an interface for connecting an external input device.

[0020] The storage unit 14 includes one or more memories. The memories may be, for example, semiconductor memories, magnetic memories, optical memories, or the like, but are not limited to these. Each memory included in the storage unit 14 may function, for example, as a main storage device, an auxiliary storage device, or a cache memory. The storage unit 14 stores any information used in the operation of the terminal device 10. Furthermore, for example, the storage unit 14 may store system programs, application programs, embedded software, and the like. For example, the information stored in the storage unit 14 may be updatable with information obtained from the network 30 via the communication unit 11, for example.

[0021] The control unit 15 includes one or more processors, one or more programmable circuits, one or more dedicated circuits, or a combination thereof. The processor may be, for example, a general-purpose processor such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit), or a dedicated processor specialized for a specific process, but is not limited to these. The programmable circuit may be, for example, but is not limited to, an FPGA (Field-Programmable Gate Array). The dedicated circuit may be, for example, but is not limited to, an ASIC (Application Specific Integrated Circuit). The control unit 15 controls the operation of the terminal device 10.

[0022] (Configuration of information processing device) As shown in FIG. 3, the information processing device 20 includes a communication unit 21, an output unit 22, an input unit 23, a storage unit 24, and a control unit 25.

[0023] The communication unit 21 includes one or more communication interfaces connected to the network 30. The communication interfaces correspond to, for example, a mobile communication standard, a wired LAN (Local Area Network) standard, or a wireless LAN standard, but are not limited to these and may correspond to any communication standard. In this embodiment, the information processing device 20 communicates with each terminal device 10 via the communication unit 21 and the network 30.

[0024] The output unit 22 includes one or more output devices that output information. Examples of the output devices include, but are not limited to, a display that outputs information as a video or a speaker that outputs information as a sound. Alternatively, the output unit 22 may include an interface for connecting an external output device.

[0025] The input unit 23 includes one or more input devices that detect an input operation by a user. Examples of the input devices include, but are not limited to, physical keys, capacitive keys, a mouse, a touch panel, a touch screen integrated with the display of the output unit 22, or a microphone. Alternatively, the input unit 23 may include an interface for connecting an external input device.

[0026] The storage unit 24 includes one or more memories. Each memory included in the storage unit 24 may function as, for example, a main storage device, an auxiliary storage device, or a cache memory. The storage unit 24 stores any information used in the operation of the information processing device 20. For example, the storage unit 24 may store a system program, an application program, embedded software, etc. In this embodiment, the storage unit 24 stores an application program required as an AI platform. In this embodiment, the storage unit 24 also stores material content that can be used for training the generating AI. The storage unit 24 also stores the generating AI (the model itself) to be provided to the user.

[0027] The control unit 25 includes one or more processors, one or more programmable circuits, one or more dedicated circuits, or a combination thereof. The processor may be, for example, a general-purpose processor such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit), or a dedicated processor specialized for a specific process, but is not limited to these. The programmable circuit may be, for example, but is not limited to, an FPGA (Field-Programmable Gate Array). The dedicated circuit may be, for example, but is not limited to, an ASIC (Application Specific Integrated Circuit). The control unit 25 controls the overall operation of the information processing device 20.

[0028] (Operation flow of information processing device) The operation of information processing device 20 according to this embodiment will be described with reference to Figures 4 and 5. Note that Figure 4 shows a first operation for training a first classification AI, and Figure 5 shows a second operation for inferring a first probability of a user's AI-generated content.

[0029] S100: The control unit 25 of the information processing device 20 stores each of a plurality of pieces of content provided by a plurality of creators as material content that can be used for training the generation AI.

[0030] Specifically, each creator uploads content to be provided as material content to the AI ​​platform to the information processing device 20 using the input unit 13 of the terminal device 10. The control unit 25 of the information processing device 20 stores each of the multiple pieces of content uploaded by multiple creators in the storage unit 24 as material content that can be used for training the generation AI.

[0031] S101: The control unit 25 assigns a label including a content ID or a creator ID to each of a plurality of material contents.

[0032] Specifically, the control unit 25 assigns a label including a content ID or creator ID designated by the administrator of the AI ​​platform to each of the multiple material contents.

[0033] The "labels" are used to train the first classification AI with the material content through supervised learning in S102. The labels may include, but are not limited to, a content ID or creator ID as described above, and may include any information that can be used to train the first classification AI.

[0034] The "content ID" is information that uniquely identifies each of the multiple material contents stored in the storage unit 24. The content ID may be expressed by alphanumeric characters or symbols, but may also be expressed by any expression, including but not limited to these. When material contents are stored in the storage unit 24, the control unit 25 may automatically determine a content ID and assign a label including the content ID to the material content.

[0035] The "creator ID" is information that uniquely identifies the creator who provided each of the multiple pieces of material content stored in the storage unit 24. The creator ID may be expressed using alphanumeric characters or symbols, but may also be expressed using any other expression, including but not limited to these. When material content is stored in the storage unit 24, the control unit 25 may identify the creator ID who provided the material content and assign a label including the identified creator ID to the material content. Furthermore, when one piece of material content is provided by multiple creators, the label assigned to the one piece of material content may include multiple creator IDs.

[0036] S102: The control unit 25 trains a first classification AI through supervised learning using the plurality of material contents as learning data, which outputs a first probability that the input content belongs to each class classified by content ID or creator ID.

[0037] Specifically, the control unit 25 trains the first classification AI by supervised learning using the plurality of material contents to which content IDs or creator IDs have been assigned in S101. Here, the first classification AI is trained to output a first probability that the input content belongs to each class classified by the content ID or creator ID.

[0038] The "first probability" output by the first classification AI trained in this way represents, for example, the degree of similarity between the input content and each piece of material content or the style of its creator used to train the first classification AI. For example, if the first probability that the content input to the first classification AI belongs to a particular class is inferred to be "80%," this means that the degree of similarity between the content and the material content indicated by the content ID corresponding to that class or the style of the creator indicated by the creator ID is inferred to be "80%."

[0039] Next, a second operation of the information processing device 20 according to this embodiment will be described with reference to FIG.

[0040] S200: Control unit 25 of information processing device 20 infers, using a first classification AI, a first probability that AI-generated content generated by a user using a generation AI belongs to each class.

[0041] Specifically, when a user inputs a desired prompt on a web page, control unit 15 of terminal device 10 transmits the prompt to information processing device 20. Control unit 25 of information processing device 20 uses a generation AI to generate AI-generated content based on the prompt. Control unit 25 then infers a first probability that the AI-generated content belongs to each class using a first classification AI.

[0042] S201: The control unit 25 determines the reward to be given to each of the multiple creators based on the inferred first probability.

[0043] For example, the control unit 25 may determine to grant a reward to the creator of the content indicated by the content ID with the highest inferred first probability, or to the creator indicated by the creator ID with the highest inferred first probability. Alternatively, the control unit 25 may determine to grant a reward to the creator of the content indicated by each content ID whose inferred first probability is equal to or greater than a threshold, or to the creator indicated by each creator ID whose inferred first probability is equal to or greater than a threshold. Here, the control unit 25 may determine to grant a larger reward to the creator with a higher inferred first probability.

[0044] The above-mentioned "threshold" used for comparison with the first probability may be determined in advance, or may be changeable by the administrator of the AI ​​platform.

[0045] The "reward" may be, for example, money, electronic money, virtual currency, or points equivalent to money, but is not limited to these and may be any reward for the creator.

[0046] As described above, the information processing device 20 stores each of the multiple pieces of content provided by multiple creators as material content available for training the generation AI. Next, the information processing device 20 assigns a label including a content ID or a creator ID to each of the multiple pieces of material content. Next, the information processing device 20 trains a first classification AI, which outputs a first probability that the input content belongs to each class classified by the content ID or creator ID, through supervised learning using the multiple material contents as training data. The information processing device 20 then infers, using the first classification AI, a first probability that the AI-generated content generated by the user using the generation AI belongs to each class. The information processing device 20 then determines a reward to be granted to each of the multiple creators based on the inferred first probability.

[0047] According to this embodiment, a first classification AI is trained through supervised learning using multiple pieces of material content assigned a content ID or creator ID as training data. This first classification AI then infers a first probability that the AI-generated content belongs to each class classified by the content ID or creator ID, and determines a reward to be granted to each of multiple creators based on the inferred first probability. Therefore, for example, a portion of the profits generated by the sale of the AI-generated content can be quantitatively and accurately distributed to rights holders according to the contribution of the material content to the AI-generated content. Therefore, according to this embodiment, technology used for rights management of AI-generated content is improved in that machine learning can quantitatively and accurately distribute profits from the AI-generated content to rights holders according to the contribution of the material content to the AI-generated content.

[0048] Although the present disclosure has been described based on the drawings and examples, it should be noted that those skilled in the art may make various modifications and alterations based on the present disclosure. Therefore, it should be noted that these modifications and alterations are included in the scope of the present disclosure. For example, the functions included in each component or step can be rearranged so as not to be logically inconsistent, and multiple components or steps can be combined or divided into one.

[0049] For example, in the above-described embodiment, the configuration and operation of the information processing device 20 may be distributed among multiple computers that can communicate with each other. Also, for example, an embodiment in which some or all of the components of the information processing device 20 are provided in the terminal device 10 may be possible. For example, the terminal device 10 may include some or all of the components of the information processing device 20.

[0050] Furthermore, for example, in the above-described embodiment, control unit 25 may store the content attributes of each of the plurality of material contents in storage unit 24, and may store a plurality of first classification AIs corresponding to the plurality of content attributes in storage unit 24. Control unit 25 may also store a second classification AI trained to output the content attributes of input content in storage unit 24. In this case, control unit 25 may train each of the plurality of first classification AIs by supervised learning using, as learning data, two or more material contents among the plurality of material contents that have content attributes corresponding to the first classification AI. Control unit 25 may also infer the content attributes of the AI-generated content using the second classification AI in S200, and infer a first probability that the AI-generated content belongs to each class using the first classification AI corresponding to the inferred content attribute.

[0051] "Content attribute" is information indicating any attribute according to the characteristics of the material content. For example, attributes of material content that is an image include attributes such as "portrait," "landscape," "realism," or "impressionism," but the attributes of material content are not limited to these examples. Also, one material content may have multiple content attributes. For example, in a certain material content, the first content attribute may be "portrait," and the second content attribute may be "realism." Also, for example, in a certain material content, the first content attribute may be "landscape," and the second content attribute may be "impressionism."

[0052] In this way, by first inferring the content attributes of input AI-generated content using the second classification AI, and then inferring the first probability that the AI-generated content belongs to each class using the first classification AI corresponding to the inferred content attributes, it is possible to reduce the amount of calculations performed by control unit 25. As a result, it is also possible to shorten the time required to infer the first probabilities.

[0053] Also, for example, in the above-described embodiment, control unit 25 may provide an AI platform on which users can sell AI-generated content.

[0054] In this way, by providing a platform where users can sell AI-generated content, the likelihood of an increase in the number of users using the AI ​​platform according to this embodiment may increase.

[0055] Furthermore, for example, in the above-described modified example, control unit 25 may determine the range within which the AI-generated content can be used, depending on the inferred first probability.

[0056] The "scope within which AI-generated content can be used" may be, for example, whether it can be sold on the AI ​​platform, the period during which it can be sold, the area in which it can be sold, or the person to whom it can be sold, but it may also be used within any scope without being limited to these.

[0057] For example, the control unit 25 may narrow the scope in which the AI-generated content can be used as the inferred first probability increases. Specifically, if the first probability that the AI-generated content belongs to a class corresponding to a certain content ID or creator ID is equal to or greater than a predetermined threshold, the AI-generated content has a relatively high similarity to the material content indicated by the content ID or the style of the creator indicated by the creator ID. According to this embodiment, the scope in which the AI-generated content can be used is narrowed (for example, by prohibiting sales on the AI ​​platform, limiting the period during which the content can be sold, or limiting the area in which it can be sold), thereby reducing the occurrence of problems related to rights.

[0058] Furthermore, for example, the control unit 25 may determine the extent to which the AI-generated content can be used by a smart contract and store it in the blockchain.

[0059] In this way, the use of blockchain technology may enable secure transactions of AI-generated content.

[0060] Also, for example, in the above-described embodiment, the control unit 25 may present the inferred first probability to a predetermined number of users to determine the validity of the inferred first probability.

[0061] Specifically, the control unit 25 uses an electronic voting system to present the inferred first probability to a predetermined number of users. If a predetermined number of users among the plurality of users judge the first probability to be valid, the control unit 25 determines that the first probability is valid. Note that the control unit 25 may execute S201 only when the first probability is judged to be valid.

[0062] The "predetermined plurality of users" refers to users who have voting rights. The predetermined plurality of users may be, for example, users who have used the AI ​​platform according to this embodiment for a certain period of time or more, users who have generated a certain amount of sales from AI-generated content, users other than the user who generated the AI-generated content, or a combination of these. However, the predetermined plurality of users may be determined based on any conditions, including but not limited to these. Note that the conditions for the predetermined plurality of users may be changeable.

[0063] The "predetermined number" may be, for example, a percentage such as a majority of a predetermined number of users, or may be a numerical value such as 10 or more. Note that the predetermined number may be changeable.

[0064] In this way, by having a portion of people with voting rights confirm the first probability (i.e., the similarity between the AI-generated content and each material content or each creator's style), the occurrence of inconveniences such as not awarding rewards to the appropriate creator because the inference accuracy of the first probability is relatively low is reduced, and rewards to creators can be awarded more accurately.

[0065] Also, for example, an embodiment is possible in which a general-purpose computer functions as the information processing device 20 according to the above-described embodiment. Specifically, a program describing the processing content for realizing each function of the information processing device 20 according to the above-described embodiment is stored in the memory of the general-purpose computer, and the program is read and executed by a processor. Therefore, the present disclosure can also be realized as a program executable by a processor, or a non-transitory computer-readable medium storing the program. [Explanation of symbols]

[0066] 1 System 10 Terminal Equipment 11 Communications Department 12 Output section 13 Input section 14 Storage section 15 Control Unit 20 Information processing equipment 21 Communications Department 22 Output section 23 Input section 24 Memory section 25 Control Unit 30 Network

Claims

1. A method executed by an information processing device, Storing each of the plurality of pieces of content provided by the plurality of creators as material content that can be used for learning the generative AI; assigning a label including a content ID or a creator ID to each of a plurality of material contents; training a first classification AI that outputs a first probability that an input content belongs to each class classified by a content ID or a creator ID through supervised learning using the plurality of material contents as learning data; Inferring a first probability that AI-generated content generated by a user using the generation AI belongs to each of the classes using the first classification AI; determining a reward to be awarded to each of the plurality of creators based on the inferred first probability; A method comprising:

2. 10. The method of claim 1, storing content attributes of each of the plurality of material contents; storing a plurality of the first classification AIs corresponding to a plurality of content attributes, respectively; and Storing a second classification AI that outputs content attributes of the input content. Further comprising: The information processing device includes: training each of the plurality of first classification AIs by supervised learning using, as training data, two or more material contents among the plurality of material contents, each of the material contents having a content attribute corresponding to the first classification AI; A method for inferring content attributes of the AI-generated content using the second classification AI, and inferring a first probability that the AI-generated content belongs to each of the classes using the first classification AI corresponding to the inferred content attributes.

3. 10. The method of claim 1, providing a platform on which the user can sell the AI-generated content.

4. 10. The method of claim 1, The method further comprising presenting the inferred first probability to a predetermined number of users to determine the validity of the inferred first probability.

5. 4. The method of claim 3, The method further includes determining an extent to which the AI-generated content can be used in response to the inferred first probability.

Images