A system that allows copyright holders to determine whether or not their copyrighted works have been studied using AI.

Abstract

Currently, during the development of various AI systems, a large amount of data and copyrighted material are used for AI training or input, regardless of whether or not the owners of the data and copyrighted material have given their consent. However, there has been no system in place for copyright holders of data and copyrighted material to determine whether or not a designated work has been trained by an AI, or to provide the necessary information for such determination. [Solution] The system stores "combinations of multiple" individual elements or keywords that are included in and considered characteristic of the copyrighted work. It then asks a designated artificial intelligence or AI a question or engages in a conversation that includes the specific keyword, and receives an answer or response from the AI ​​that includes the keyword. The system then calculates the probability that the AI's response contains "another individual keyword" that is stored in combination with the "specific keyword".

Description

Technical Field

[0001] The present invention relates to a technology related to an artificial intelligence or AI (hereinafter referred to as "AI") system that features learning a large number of data and works.

Background Art

[0002] Currently, when developing various AIs, a large number of data and works are learned by AI or input into AI, regardless of whether the owners of specific data and works have given consent.

[0003] There has been no system that allows copyright holders who are not involved in the development of AI itself to determine whether a specific AI has learned specific data or works.

Prior Art Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] The technology of the "transfer management system for works" in Patent Document 1 was a system that enabled contracts and transfers of rights for a vast number of works and copyrights. However, there was no system that could calculate the probability that a specified AI had learned specific copyrights or data and was useful for determining whether it had learned them.

Means for Solving the Problems

[0006] The invention according to claim 1 is a system that asks questions or has conversations with a specified AI and calculates the probability that the answer from the AI contains "another individual keyword" that is memorized in combination with a "specific keyword included in the question".

[0007] According to the present invention, it is possible to calculate the probability that a specified AI has learned specific data, and thereby determine whether or not a specified AI has learned specific data using probability.

[0008] The invention of claim 2 enables the technology of the invention of claim 1 to be performed by a system such as AI, or by a person using voice or data input, or through dialogue. [Effects of the Invention]

[0009] This invention provides authors or copyright holders (hereinafter collectively referred to as "copyright holders") of various copyrighted works (hereinafter referred to as "works") described or expressed in specific data, media, or publications including electronic publications, with criteria for determining whether their copyrights or data have been learned by a designated AI. This allows copyright holders who are not involved in the development of the AI ​​itself to determine whether a designated AI has learned specific data. As a result, we believe that the technology of this invention will contribute to sound AI development in the future by helping to protect the ownership rights of specific data and copyrighted works in AI development. Furthermore, when copyright holders of copyrighted works determine whether an AI has learned their copyrighted works, they do not need to disclose or transfer their copyrighted works to any party other than this system. Also, when determining or investigating whether an AI has learned a work, it is not necessary to investigate the internal workings of the AI ​​system or to obtain the cooperation of those involved in its development.

[0010] Furthermore, the guidelines on "AI and Copyright" issued by the Agency for Cultural Affairs in 2024 state that, as a requirement for a copyrighted work (such as an answer or reply) generated independently by AI to infringe on the copyright of a third party's work, "whether the essential characteristics of the expression can be directly perceived" is included. The technology of the present invention can provide a numerical value (%) for the similarity rate of essential characteristics of expression, which can be used as material for this determination. [Brief explanation of the drawing]

[0011] [Figure 1] A system that allows copyright holders to determine whether or not their copyrighted works have been studied using AI. [Modes for carrying out the invention]

[0012] An example of an embodiment of the present invention will be described with reference to Figure 1 and using the language of Claim 1. The system (1) that allows copyright holders to determine whether or not their copyrighted works have been learned by AI comprises a storage device (2) for recording and storing information, an information processing means (3) for interacting with a designated AI and analyzing information from conversations or question-and-answer sessions, and an information transmission and reception means (4) for sending and receiving information. It is also capable of communicating via a network such as the Internet with a designated AI or an information terminal (5) related to the designated AI that learns data and copyrights.

[0013] Furthermore, the memory device (2) stores "keyword A and keyword B," which are individual elements or keywords described or expressed in a specific data or media copyrighted work, linked together. It is also advisable to store a large number of related or linked individual elements or keywords (for example, keyword C and keyword D) within the individual elements or keywords described or expressed in a specific data or media copyrighted work, similar to keyword A and keyword B. These individual elements or keywords are data related to text, audio, music, images, or video.

[0014] The information processing begins with the information processing means (3) creating a "keyword A question" using keyword A stored in the memory device (2), and then transmitting it from the information transmission / reception means (4) to the information terminal (5) via a network such as the Internet.

[0015] Next, the "answer or response to the question about keyword A" generated or created by the designated AI, which is transmitted from the information terminal (5), is received by the information transmission / reception means (4) via a network such as the Internet, and received by the information processing means (3).

[0016] The information processing means (3) refers to the storage device (2) to determine whether the received "answer or reply to the question for keyword A" contains data that includes keyword B, which is stored in association with keyword A (hereinafter referred to as "data containing keyword B"), and stores the result of that determination in the storage device (2).

[0017] Subsequently, the information processing means (3) divides the number of data stored in the memory device (2) that have been determined to contain "keyword B" by the number of "answers or responses to questions about keyword A" that have been performed, and calculates the probability (hereinafter referred to as "the probability") that "data containing keyword B" is included in the answers or responses produced or created by the designated AI. The same information processing is also performed using "other combinations" of keyword C and keyword D relating to the same specific data or copyrighted work such as media, and these are included in the calculation of the probabilities relating to keyword A and keyword B to calculate the combined probability of the question-and-answer sessions performed. The calculated probability is then transmitted to a designated email address or other recipient via the information transmission means (4).

[0018] Furthermore, the memory device (2) may be configured to store a specified probability (numerical value), and if the probability calculated by the information processing means (3) exceeds the specified probability (numerical value), the system may be configured to send a "determination of whether the specified AI has learned the specific data" to a specified email address or other recipient via the information sending / receiving means (4).

[0019] Furthermore, the sending and receiving of "keyword A questions" and "answers or responses to keyword A questions" generated or created by the designated AI may be done not through data communication via the internet or other networks, but rather through human interaction using voice or data input, or through dialogue, and this may be done by connecting equipment or systems capable of such interaction to this system (1). [Industrial applicability]

[0020] The technology of the present invention is considered to contribute to the sound development of AI in the future by helping to protect the ownership of specific data and works in AI development.

Explanation of Signs

[0021] System (1) for determining whether a specified AI has learned specific data Storage device (2), Information processing means (3), Information transmission / reception means (4) Specified AI or information terminal (5) related to the specified AI

Claims

1. This system allows copyright holders to determine whether or not their copyrighted works have been learned by AI. It features a mechanism where the author or copyright holder (hereinafter collectively referred to as "copyright holder") of various copyrighted works (hereinafter referred to as "works") contained in or expressed in specific data, media, or publications including electronic publications, memorizes "multiple combinations" of individual elements or keywords (hereinafter collectively referred to as "individual keywords") that they consider to be included in and characteristic of the works. The copyright holder then asks a designated artificial intelligence or AI (hereinafter collectively referred to as "AI") a question or conversation (hereinafter referred to as "keyword-containing question") that includes the specific individual keywords, and receives an answer or response (hereinafter referred to as "response from AI") from the AI. The system then calculates the probability that the AI's response contains "another individual keyword" that has been memorized in combination with "a specific individual keyword included in the keyword-containing question."

2. A system described in claim 1 that allows the copyright holder to determine whether or not the copyrighted work has been learned by AI, characterized in that questions including keywords or answers from AI are provided by an AI or similar system, or that a person inputs data or engages in voice conversation and records or inputs them as data.

Images