Method for preventing outflow of idea originality in automatic program generation function

Abstract

[Problem] When generative AI and automatic programming that does not require human coding become mainstream, code created by automatic program generation functions using similar natural language inputs will become similar, and it will be easy for others to imitate the code. The objective of the present invention is to realize automatic programming that guarantees originality. [Solution] Sentences or phrases input into an automatic program generation function are recorded, and if an equivalent sentence or phrase is subsequently input, program execution with respect to the latter input is rendered impossible, thereby preventing the generation of programs generated by inputting equivalent sentences or phrases.

Description

Methods to prevent the leakage of originality of ideas in automated program generation functions

[0001] This invention relates to an automatic program generation function for automatically creating software, also known as automatic programming, and more particularly to a technology that is effective when applied to an environment for executing multiple automatic programs, i.e., an automatic program generation function.

[0002] As generative AI becomes more widely used, and automatic program generation functions that do not require human programming—so-called automatic programming—become mainstream, it is thought that programs will be generated by artificial intelligence using natural language processing, and human programming will decrease. Currently, automatic programming generates programs using a system that associates natural language with code created by humans. The programming code for natural language is associated with a database linked to the automatic program generation function, or within the program that has the automatic program generation function. Code corresponding to new words and sentences is added. The destination for this code is within the program with the automatic program generation function or within a commonly used database. As a result, regardless of the natural language input by each user, the system learns and grows in a common way, and all users can utilize the relationship between the learned code and natural language. The structure of having natural language-corresponding code in this database is synchronized sequentially even if there are multiple databases, and the same database is used. Furthermore, the structure of having natural language-corresponding code within the automatic program generation function allows for the maintenance of an identical automatic program generator by adding the learned results to all automatic program generation functions, recompiling, and synchronizing sequentially. This allows all users to similarly enjoy the learned results.

[0003] On the other hand, a disadvantage of the automatic program generation function under this same environment is that using a scenario with the same natural language sentence enables the generation of an identical program. Even if application software with added value is developed by realizing a unique idea, providing a similar natural language sentence allows the development of a similar analogous application software, which gives rise to the problem of loss of competitiveness. Furthermore, since an automatic program generation function like generative AI has a learning function, when a program created from a natural language sentence provided to one automatic program generation function is shared among all automatic program generation functions, loss or leakage of the uniqueness of ideas may occur.

[0004] Japanese Unexamined Patent Publication No. 09-073306: An input method for program creation that is made closer to natural language (several types are proposed), and a database input method closely related to program input Japanese Unexamined Patent Publication No. 06-214776: One that automatically generates a target program source by inputting software specifications written in natural language Japanese Patent No. 7064680: One obtained by adding a natural language processing method to Japanese Unexamined Patent Publication No. 06-214776 and storing corresponding program codes in a database Japanese Patent No. 6958884: One that has the same content as Japanese Unexamined Patent Publication No. 06-21477 and adds conversation reception Japanese Patent Application No. 2018-212991: A method of generating feedback information associated with input content and providing the same to a user terminal Japanese Patent Application No. 2012-88951: A method for reflecting correction content in a newly automatically generated program when a change is made to a program generated by automatic programming Japanese Patent Application No. Sho 62-39145: An artificial intelligence that automatically creates a program by providing natural language to artificial intelligence and adding minimal human assistance Japanese Patent No. 7573141: Idea leakage prevention during automatic programming Japanese Patent No. 7575736: Idea leakage prevention during automatic programming

[0005] In systems and programs with automatic program generation capabilities, including generative AI, using the same automatic program generation function in different environments will result in the generation of identical program code by inputting the same natural language sentence. This means that unique ideas in the input natural language sentence can leak out through the generated identical program code. For example, it could easily generate programs with similar functionality by describing the functionality of existing applications, or it could lead to problems such as the creation of identical programs for cryptographic techniques, which could then be used to decrypt them.

[0006] A system with an automatic program generation function, or the program generation function itself, memorizes the input text or phrase, and if the same text or phrase is input a second time or later, it generates a program that cannot be executed. By preventing the execution of the generated program for similar text or phrase inputs, it prevents the leakage of ideas.

[0007] Examples of typical artificial intelligence workflows, including generative AI, automatic program generation functions, and so-called automatic programming, where natural language processing divides given content and retrieves corresponding code from a database; examples of code retrieval when using generative AI; examples of program generation functions that have the content of input text or phrases and allow the program to be executed if the content is equivalent; examples of workflows that allow / deny execution of a program if another person makes similar input after the user's input; examples of workflows that allow execution of equivalent programs when generating a program from text input.

[0008] The embodiments of the present invention will now be described in detail with reference to the drawings. However, the present invention can be implemented in many different forms and should not be interpreted as being limited to the contents of these embodiments. The same elements will be numbered throughout the embodiments.

[0009] The following embodiments primarily describe methods or systems, but as will be apparent to those skilled in the art, the present invention can also be implemented as a computer program. Therefore, the present invention can take the form of hardware, software, or a combination of software and hardware. The program can be recorded on any computer-readable medium such as a hard disk, DVD-ROM, optical or magnetic storage device, or semiconductor storage device.

[0010] Furthermore, in the following embodiments, general computer systems and the like can be used. The computer systems, network systems, and artificial intelligence systems that can be used in these embodiments are equipped with hardware resources that are generally found in computer systems, network systems, and artificial intelligence systems, such as a central processing unit (CPU), main memory (RAM), non-volatile memory (ROM), coprocessor, image accelerator (GPU), AI semiconductors, cache memory, input / output control devices (I / O), semiconductors for artificial intelligence functions, and flash memory. They can also be equipped with external storage devices such as hard disk drives and flash memory, and communication means that can connect to a network such as the Internet. Computer systems include various types of computers such as personal computers, workstations, mainframe computers, and servers, as well as cloud-based network-connected computers, tablet computers, smartphones and other mobile information terminals, and systems that constitute artificial intelligence.

[0011] The present invention aims to prevent similar program code from being output even with highly unique input content by adding the program code generated by the automatic program generation function to the automatic program generation function, thereby enabling the learning and growth of the generative AI or automatic program generation function to output similar programming code when similar natural language sentences are input. In other words, it aims to prevent the breach of confidentiality and the leakage of uniqueness.

[0012] Generative AI and other automated programming functions generally divide the input text into phrases, as shown in Figure 1. Program elements corresponding to each phrase are extracted from a database or similar source, and combined according to arbitrary or rule-based procedures to generate a program. The results are then further combined to generate a program that meets the intended purpose.

[0013] As shown in Figure 2, the present invention comprises an automatic program generation function that includes a function to hold input text or phrases separated from input text, and a permission flag associated with a user ID that, when equivalent input text or phrases are received after the holding period, sets whether program execution is permitted or denied within the program generated for that input text or phrase. Optionally, by setting a permission period for the user ID, it becomes possible to set a period during which execution is permitted or denied.

[0014] Figure 3 shows an example of a flow where, at any time, the user can set whether to allow or deny the execution of the program when it is generated, based on whether the input text, divided phrases, or combinations of such phrases are entered after the user's input text. If the user who initially entered the text sets it to allow execution, then even if another person enters equivalent input, that person will be able to set whether the program is executable or not when the program is automatically generated. For example, if it is set to deny execution, and equivalent content is already saved, then even if another person enters equivalent content, when that person performs automatic program generation, an executable program will be generated.

[0015] Figure 4 shows an example of a simplified flow for the program generation prevention function. For automatic program generation, a text describing the expected function is input. In generation AI and automatic programming, the input text is generally processed using natural language processing, for example, being divided into sentences or phrases that constitute unit functions, and then further divided by content. The system checks whether a sentence or phrase with equivalent content exists in a data storage device, such as a database, that was saved before this input. In other words, a comparison with saved content is performed. If no sentence or phrase with equivalent content exists in the saved content, this content is saved as new content along with the inputter's user ID. Furthermore, the automatic program generation function generates a program based on the input content. As a result, a program is generated. If the comparison with saved content reveals that equivalent content already exists in the saved data, the program execution permission flag associated with the user ID that initially entered the saved data is checked. If the program execution permission flag is granted, the program execution permission is set within the program based on the input content, and automatic program generation is performed. If the program execution permission flag is denied, the program automatic generation function will set a program execution denial setting within the program based on the input content and perform automatic program generation. As a result, the program will be generated by the program automatic generation function. If the permission flag is denied, the program will be denied and the user will not be able to execute it. As a result, even if there is equivalent input, the generation of equivalent programs can be prevented. However, if the user who initially entered the text grants permission to subsequent user IDs who entered equivalent text, it will be possible to generate an executable program. Also, if the user who initially entered the text grants execution permission to all subsequent inputs, a normal executable program will be generated.

[0016] This technology can be applied to generative AI, enabling the creation of unique application software tailored to each individual or company to prevent leakage to others.

[0017] Furthermore, it can be used on websites that sell proprietary program code for a fee, and by applying it to a generation AI that provides pre-created function code only to registered users, it becomes possible to grant permission for proprietary know-how to be used only by specific users.

[0018] Furthermore, by adding settings such as the duration for which the program can be executed, it becomes possible to generate programs that can only be executed during the licensed period, or programs that cannot be executed until a certain period has passed.

Claims

1. A system equipped with a generation AI or automatic program generation function that generates a program containing errors when a second or subsequent combination of sentences or phrases with the same content as the first input combination of sentences or phrases is entered.

2. A system according to claim 1, comprising a generation AI or automatic program generation function that generates an error-free program when the first inputter grants permission to subsequent inputters.

Images