Method for integrally using artificial intelligence multi-model

Abstract

The present invention relates to a method for integrally using an artificial intelligence multi-model, comprising: a model list output step (S100); an options transmission step (S200); a preprocessing step (S300) for preprocessing data received from a server (10) after the options transmission step (S200); an execution variable setting step (S400) in which the server (10) sets an execution variable of an artificial intelligence model after the preprocessing step (S300); a model execution step (S500) in which the server (10) executes the selected artificial intelligence model on the basis of data input by a user so as to acquire a result value after the execution variable setting step (S400); and a result output step (S600) in which the server (10) outputs the acquired result value to a user terminal (20) after the model execution step (S500).

Description

How to use AI multi-model integration

[0001] The present invention relates to a method for integrating and using multiple artificial intelligence models. More specifically, it relates to a method designed to enable users to more easily utilize various types of artificial intelligence models.

[0002] The artificial intelligence market has experienced explosive growth in recent years, and this rapid growth has led numerous companies to jump into AI technology development. To remain competitive, these companies are developing individual AI models in their own unique ways, and are making some of their results available for limited use in limited spaces. Considering the nature of AI, where the combination of diverse data and AI models can lead to the emergence of diverse services and enhanced model performance, it's difficult to see this closed environment as having a positive impact on market growth. Furthermore, while AI models are inherently useful, the difficulty of training and utilizing them hinders their accessibility. The method for integrating multiple AI models according to the present invention is designed to enable users to more easily utilize a variety of AI models. Related literature is presented below.

[0003] Prior art documents 001 to 004 are inventions having technical relevance to the present invention, and prior art document 001 relates to an artificial intelligence vulnerability diagnosis learning recommendation service system and method, prior art document 002 relates to a method for automatically analyzing fine particles distributed over a large area using artificial intelligence, an electron microscope, and an EDS analyzer, prior art document 003 relates to a game quality assurance method and system using an artificial intelligence agent, and prior art document 004 relates to an artificial intelligence-based brain disease diagnosis method and system using human dynamic characteristic information.

[0004] Prior art documents 001 to 004 are all related to artificial intelligence, and thus have similarities with the method for integrating artificial intelligence multi-models according to the present invention. However, prior art documents 001 to 004 differ from the present invention in that they do not use multiple artificial intelligences.

[0005] The present invention relates to a method for integrating and using artificial intelligence multi-models.

[0006] The present invention has been devised to solve the problems of the prior art as described above, and comprises a model list output step (S100) of transmitting a list of available artificial intelligence models from a server (10) to a user terminal (20) and outputting the list of artificial intelligence models received from the user terminal (20), a selection transmission step (S200) of transmitting the selected artificial intelligence model and the input data from the user terminal (20) to the server (10) after the model list output step (S100) when a user selects an artificial intelligence model to be used through the user terminal (20) and inputs data, a preprocessing step (S300) of preprocessing the data received from the server (10), an execution variable setting step (S400) of setting the execution variables of the artificial intelligence model from the server (10) after the preprocessing step (S300), a model execution step (S500) of executing the selected artificial intelligence model based on the data input by the user from the server (10) to obtain a result value after the execution variable setting step (S400), After the above model execution step (S500), a result output step (S600) is included for outputting the result value obtained from the server (10) to the user terminal (20).

[0007] The present invention has been devised to solve the problems of the prior art as described above, and includes an accuracy judgment step (S700) for judging the accuracy of the result value in the server (10) before the result output step (S600) and after the model execution step (S500).

[0008] The present invention has been devised to solve the problems of the prior art as described above, and the number of artificial intelligence models included in the list of available artificial intelligence models transmitted from the server (10) to the user terminal (20) in the model list output step (S100) is at least two, the user selects at least two artificial intelligence models to be used through the user terminal (20), and includes an integrated method candidate transmission step (S310) for transmitting an artificial intelligence model integration method candidate from the server (10) to the user terminal (20) after the preprocessing step (S300) and before the execution variable setting step (S400).

[0009] The present invention has been devised to solve the problems of the prior art as described above, and the model list output step (S100) outputs sample data on the expected results when using the artificial intelligence model selected by the user together with the artificial intelligence model list.

[0010] The present invention has been devised to solve the problems of the prior art as described above, and the model list output step (S100) outputs sample data on expected results when a plurality of artificial intelligence models selected by a user are combined and used together with the artificial intelligence model list.

[0011] According to the method of integrating artificial intelligence multi-models according to various embodiments of the present invention as described above, there is an effect of obtaining more accurate results because multiple artificial intelligence models are used.

[0012] In addition, according to the present invention, if it is determined that the result value has an accuracy below a certain level in the accuracy judgment step, the model variables are reset and the artificial intelligence model is re-executed, so there is an effect of ensuring an accuracy above a certain level.

[0013] In addition, according to the present invention, since sample data for expected results when a specific artificial intelligence model or a combination of multiple artificial intelligence models is output to the user terminal at the model list output stage, there is an effect in which the user can select a more suitable artificial intelligence model.

[0014] Figure 1 is a block diagram of a user terminal and a server that perform an artificial intelligence multi-model integrated use method according to the present invention.

[0015] Figure 2 is a block diagram of a user terminal, server, and database that perform an artificial intelligence multi-model integrated use method according to the present invention.

[0016] Figure 3 is a flowchart from a user perspective of the artificial intelligence multi-model integration method according to the present invention.

[0017] Figure 4 is a flowchart of the artificial intelligence multi-model integrated use method according to the present invention from the perspectives of a user terminal and server.

[0018] Figures 5 to 10 are flowcharts of various embodiments of an artificial intelligence multi-model integrated use method according to the present invention.

[0019] Hereinafter, the most preferred embodiment of the present invention will be described in detail to enable a person having ordinary skill in the art to easily carry out the present invention.

[0020] The numbers cited in the examples below are not limited to the referenced objects and can be applied to all examples. Objects that achieve the same purpose and effect as the configurations presented in the examples are equivalent replacement objects. The superordinate concepts presented in the examples include subordinate concepts not described.

[0021]

[0022] [Embodiment 1-1] The present invention relates to a method for integrating and using artificial intelligence multi-models, comprising: a model list output step (S100) for transmitting a list of available artificial intelligence models from a server (10) to a user terminal (20) and outputting the list of artificial intelligence models received from the user terminal (20); a selection transmission step (S200) for allowing a user to select an artificial intelligence model to be used through the user terminal (20) and input data, and transmitting the selected artificial intelligence model and input data from the user terminal (20) to the server (10); a preprocessing step (S300) for preprocessing data received from the server (10) after the selection transmission step (S200); an execution variable setting step (S400) for setting execution variables of the artificial intelligence model from the server (10) after the preprocessing step (S300); and a model for obtaining a result value by executing the selected artificial intelligence model based on data input by the user from the server (10) after the execution variable setting step (S400). The execution step (S500) includes a result output step (S600) for outputting the result value obtained from the server (10) to the user terminal (20) after the model execution step (S500).

[0023] [Example 1-2] The present invention relates to a method for integrating artificial intelligence multi-models, and in Example 1-1, after the model execution step (S500), includes an update step (S510) for storing the result value in a database (30).

[0024] [Example 1-3] The present invention relates to a method for integrating artificial intelligence multi-models, and in Example 1-1, after the optional transmission step (S200), a sample data determination step (S210) is included for determining whether data input by the user in the server (10) is identical to pre-stored sample data.

[0025] [Example 1-4] The present invention relates to an integrated use method of artificial intelligence multi-models, and in Example 1-3, after the sample data determination step (S210), if it is determined that the data input by the user is identical to the pre-stored sample data, a sample data transmission step (S220) is included for transmitting the pre-stored sample data from the server (10) to the user terminal (20).

[0026] [Example 1-5] The present invention relates to a method for integrating artificial intelligence multi-models. In Example 1-3, if it is determined in the sample data determination step (S210) that the data input by the user is not identical to the pre-stored sample data, the preprocessing step (S300) is performed.

[0027] [Example 1-6] The present invention relates to a method for integrating artificial intelligence multi-models. In Example 1-3, the sample data is stored in a database (30).

[0028] [Example 1-7] The present invention relates to a method for integrating and using artificial intelligence multi-models. In Example 1-1, the model list output step (S100) transmits the artificial intelligence model list together with characteristic information for each artificial intelligence model from the server (10) to the user terminal (20), and the user terminal (20) outputs the characteristic information for each artificial intelligence model received from the server (10).

[0029] The present invention relates to a method for integrating and utilizing multiple AI models. The AI ​​market has experienced explosive growth in recent years, and this rapid growth has led numerous companies to jump into AI technology development. To remain competitive, these companies are developing individual AI models in different ways and making some of the results available for limited use in limited spaces. Considering the nature of AI, where the combination of diverse data and AI models can lead to the emergence of diverse services and enhanced model performance, it's difficult to see this closed environment as having a positive impact on market growth. Furthermore, while AI models are inherently useful, the difficulty of training and utilizing them makes them difficult for many people to access. The method for integrating multiple AI models according to the present invention is designed to enable users to more easily utilize a variety of AI models.

[0030] In order to solve the technical problems described above, the artificial intelligence multi-model integrated use method according to the present invention may include a model list output step (S100), an optional transmission step (S200), a preprocessing step (S300), an execution variable setting step (S400), a model execution step (S500), and a result output step (S600).

[0031] The model list output step (S100) is a step in which a list of artificial intelligence models available to the user is transmitted from the server (10) to the user terminal (20), and the user terminal (20) outputs the list of artificial intelligence models received from the server (10). The user terminal (20) may be an electronic device that can connect to the server (10). For example, the user terminal (20) may be an electronic device such as a smartphone, a PC, or a tablet. The user terminal (20) may connect to the server (10) via a wired or wireless communication method. Here, the wireless communication method may be at least one of a mobile communication network and wireless Internet. In the model list output step (S100), the server (10) may transmit a list including at least two artificial intelligence models to the user terminal (20). The user may select a specific artificial intelligence model from the list of artificial intelligence models output to the user terminal (20). The artificial intelligence model used in the present invention may be an artificial intelligence model that is generally available to users in the market, and may be, for example, an artificial intelligence model such as ChatGPT, Gemini, or Stable Diffusion.

[0032] The optional transmission step (S200) is a step in which, when a user selects an artificial intelligence model and inputs data, the user's optional information and data of the artificial intelligence model are transmitted to the server (10) from the user terminal (20). Here, the data may be data related to a specific problem that the artificial intelligence model is to solve. The artificial intelligence model largely performs tasks such as prediction, classification, generation, and recommendation. In the tasks performed by the artificial intelligence model, the user's request may mainly be text, but in some cases, there may be requests for images, videos, and music. For example, when a user requests that two specific images be synthesized, the data that the user inputs through the user terminal (20) may be an image, and the result may also be an image. As another example, when a user requests that a specific book be recommended, the data that the user inputs through the user terminal (20) may be text, and the result may also be text.

[0033] The preprocessing step (S300) is a step in which the server (10) preprocesses data received from the user terminal (20). Preprocessing is the process of converting data into a form that the artificial intelligence model can understand so that the artificial intelligence model can understand the question and provide an accurate answer. Preprocessing is performed to ensure consistency with the model training data, remove unnecessary information, and convert the input data into a form that the artificial intelligence model can process. For example, if the data entered by the user into the user terminal (20) is in text format, the preprocessing step involves tokenizing the user-entered data into sentences, removing stop words, analyzing morphemes, vectorizing sentences, and normalizing the data. Here, sentence tokenization refers to the task of dividing sentences into word or phrase units, stop word removal refers to the task of removing words that do not add meaning, such as '의', '가', and '는', morphological analysis refers to the task of dividing words into the smallest meaningful units, sentence vectorization refers to the task of converting sentences into numeric vectors, and normalization refers to the task of converting various expression methods into a single standard form. This preprocessing process is an important step for improving the performance and generalization ability of artificial intelligence models, and different preprocessing methods are used depending on the type of artificial intelligence model.

[0034] The execution variable setting step (S400) sets the execution variables of the artificial intelligence model on the server (10). The execution variables of the artificial intelligence model refer to various setting values ​​that affect the results when the artificial intelligence model performs a specific task. Examples of execution variables of the artificial intelligence model include learning rate, epoch, batch size, activation function, optimizer, regularization, and dropout.

[0035] The model execution step (S500) executes an artificial intelligence model based on data input by the user in the server (10) to obtain a result value.

[0036] The result output step (S600) is a step in which the result value obtained through the model execution step (S500) is transmitted from the server (10) to the user terminal (20), and the user terminal (20) outputs the received result value.

[0037] The method for integrating artificial intelligence multi-models according to the present invention may include an update step (S510) performed after the model execution step (S500).

[0038] The update step (S510) stores the result value in the database (30).

[0039] The database (30) may include separate sample data. Sample data may be data regarding examples of what kind of results will be produced when a specific artificial intelligence model is used when a user checks the list of artificial intelligence models output on the user terminal (20), and may be output together with the list of artificial intelligence models. In some cases, the user may want the same kind of result value as this sample data. Therefore, the method for integrating artificial intelligence multi-models according to the present invention may further include, after the optional transmission step (S200), a sample data determination step (S210) in which the server (10) determines whether the data input by the user is identical to pre-stored sample data.

[0040] The pre-stored sample data used in the sample data determination step (S210) may be stored in the database (30). That is, in the sample data determination step (S210), the server (10) determines through mutual communication with the database (30) whether the data entered by the user matches the sample data stored in the database (30). If the data entered by the user is determined to be identical to the sample data in the sample data determination step (S210), the server (10) transmits the sample data pre-stored in the database (30) from the server (10) to the user terminal (20) through the sample data transmission step (S220), so that the sample data is output from the user terminal (20). However, if the data entered by the user in the sample data determination step (S210) is determined to be not identical to the sample data pre-stored in the database (30), the preprocessing step (S300) may be performed.

[0041] In the model list output step (S100), the user must check the list of artificial intelligence models output to the user terminal (20) and select a specific artificial intelligence model. The name of each artificial intelligence model is output in this artificial intelligence model list. However, if the name of the artificial intelligence model is simply output, it is difficult for a user who does not primarily use artificial intelligence models to know which artificial intelligence model he or she wants. For example, a user who wants to generate a specific image must select an artificial intelligence model that generates the image, but it is difficult to determine which of the artificial intelligence models output to the user terminal (20) is the model that generates the image. Therefore, in the model list output step (S100), the server (10) can transmit the characteristics of each artificial intelligence model together with the list of artificial intelligence models to the user terminal (20), and the user terminal (20) can output characteristic information for each artificial intelligence model together with the list of artificial intelligence models received from the server (10), thereby providing the user with additional information when selecting an artificial intelligence model.

[0042] In addition, unlike this method, in the model list output step (S100), a method may also be used in which the user terminal (20) provides categories of artificial intelligence models, and when the user selects a specific category, a list of artificial intelligence models corresponding to that type is output.

[0043]

[0044] [Example 2-1] The present invention relates to a method for integrating artificial intelligence multi-models, and in Example 1-1, after the model execution step (S500) and before the result output step (S600), the server (10) includes an accuracy judgment step (S700) for judging the accuracy of the result value.

[0045] [Example 2-2] The present invention relates to a method for integrating and using artificial intelligence multi-models. In Example 2-1, if the server (10) of the accuracy determination step (S700) determines that the accuracy of the result value is below a reference value, the server re-executes the model execution variable setting step (S400), the model execution step (S500), and the accuracy determination step (S700).

[0046] [Example 2-3] The present invention relates to a method for integrating artificial intelligence multi-models. In Example 2-1, if the server (10) determines in the accuracy determination step (S700) that the accuracy of the result value is higher than the reference value, the server performs the result output step (S600).

[0047] The present invention may further include an accuracy judgment step (S700). The accuracy judgment step (S700) is a step performed after the model execution step (S500), and determines the accuracy of the result value resulting from the execution of the artificial intelligence model on the server (10) before the result output step (S600).

[0048] In the accuracy judgment step (S700), if the server (10) determines that the accuracy of the execution result of the artificial intelligence model is below the standard, the server can re-execute the model execution variable setting step (S400), the model execution step (S500), and the accuracy judgment step (S700) to increase the accuracy of the result. This is because the accuracy may vary depending on the execution variables set in the model execution variable setting step (S400).

[0049] However, if the server (10) determines that the accuracy of the execution result of the artificial intelligence model exceeds the standard value in the accuracy judgment step (S700), the server may perform the result output step (S600) to output the result value to the user terminal (20).

[0050]

[0051] [Embodiment 3-1] The present invention relates to an artificial intelligence multi-model integrated use method, and in Embodiment 1-1, the number of artificial intelligence models included in the list of available artificial intelligence models transmitted by the server (10) to the user terminal (20) in the model list output step (S100) is at least two, the user selects at least two artificial intelligence models to be used through the user terminal (20), and includes an integrated method candidate transmission step (S310) for transmitting an artificial intelligence model integration method candidate from the server (10) to the user terminal (20) after the preprocessing step (S300) and before the execution variable setting step (S400).

[0052] [Embodiment 3-2] The present invention relates to a method for integrating and using artificial intelligence multi-models. In Embodiment 3-1, in the integration method candidate transmission step (S310), the artificial intelligence model integration method candidate includes at least one of a pipeline method, a merge method, a voting method, an average method, a weighting method, and a stacking method.

[0053] [Example 3-3] The present invention relates to a method for integrating and using artificial intelligence multi-models. In Example 3-1, after the integration method candidate transmission step (S310) and before the execution variable setting step (S400), if a user selects one of the artificial intelligence model integration method candidates, a model integration step (S320) is included for integrating at least two artificial intelligence models using the selected artificial intelligence model integration method.

[0054] [Example 3-4] The present invention relates to a method for integrating and using artificial intelligence multi-models. In Example 3-1, after the preprocessing step (S300) and before the integration method candidate transmission step (S310), a model integration possibility determination step (S330) is included to determine whether the artificial intelligence model selected by the user in the server (10) is a combination that can be integrated.

[0055] [Example 3-5] The present invention relates to a method for integrating artificial intelligence multi-models. In Example 3-4, in the step of determining whether model integration is possible (S330), if the server (10) determines that the artificial intelligence model selected by the user is not a combination that can be integrated, it transmits a signal to the user terminal (20) notifying that integration is impossible.

[0056] [Example 3-6] The present invention relates to a method for integrating and using artificial intelligence multi-models. In Example 3-4, in the step of determining whether model integration is possible (S330), if the server (10) determines that the artificial intelligence model selected by the user is integrable, it performs the step of transmitting the integration method candidate (S310).

[0057] The user can select multiple artificial intelligence models included in the list of artificial intelligence models output to the user terminal (20) in the model list output step (S100). That is, in the present invention, multiple artificial intelligence models can be used together to obtain better results.

[0058] In order to use multiple artificial intelligence models, a method for integrating multiple artificial intelligence models is selected, and for this purpose, the artificial intelligence multi-model integration method according to the present invention may include an integration method candidate transmission step (S310) performed between the preprocessing step (S300) and the execution variable setting step (S400).

[0059] The integration method candidate transmission step (S310) is a step of transmitting candidate information of an artificial intelligence model integration method from the server (10) to the user terminal (20). The candidate information of the artificial intelligence model integration method may include a pipeline method, a merge method, a voting method, an averaging method, a weighted averaging method, a stacking method, etc. However, the candidates of the multiple artificial intelligence model integration methods used in the present invention are not limited to those described above, and other artificial intelligence model integration methods may be used in addition to the above-described methods.

[0060] Pipelining is a method of sequentially connecting multiple AI models to perform a single, complex task. In a pipeline, each AI model is responsible for a specific subtask, and the output of the previous AI model is connected to the input of the next AI model. Pipelining is particularly advantageous for solving complex problems.

[0061] Merge refers to the process of combining multiple AI models into one, combining the parameters or output values ​​of each model to create a new model. When merging multiple AI models to create a new model, the parameters can be an average of the parameters of each model. Merge can reduce the size of the AI ​​model, reduce computational costs, and achieve higher performance by combining the strengths of individual models.

[0062] The voting method, averaging method, weighting method, and stacking method described above are integrated methods for artificial intelligence models that process the results.

[0063] When a user selects one of the artificial intelligence model integration method candidates output on a user terminal (20), a plurality of artificial intelligence models selected by the selected method are integrated in a model integration step (S320) performed after the integration method candidate transmission step (S310), and then an execution variable setting step (S400) is performed.

[0064] In some cases, the multiple AI models selected by the user in the model list output step (S100) may be a combination that is difficult to integrate with each other. For example, if the user selects Stable Diffusion, which mainly generates images, and ChatGPT, which outputs answers to questions in text format, these two AI models may not be integrated because they have different output and execution methods. Therefore, the server (10) may transmit a signal to the user terminal (20) notifying that integration is impossible if the multiple AI models selected by the user are a combination that is difficult to integrate with each other through the preprocessing step (S300) and the model integration possibility determination step (S330) in which the integration method candidates are transmitted. If the server (10) determines in the model integration possibility determination step (S330) that the AI ​​models selected by the user are a combination that is possible to integrate, the integration method candidate transmission step (S310) may be performed.

[0065]

[0066] [Example 4-1] The present invention relates to a method for integrating and using artificial intelligence multi-models. In Example 1-1, the model list output step (S100) outputs sample data on expected results when using an artificial intelligence model selected by a user, together with the artificial intelligence model list.

[0067] [Example 4-2] The present invention relates to a method for integrating artificial intelligence multi-models. In Example 4-1, the sample data is pre-stored in a database (30).

[0068] The artificial intelligence models used in the present invention each have their own characteristics. For example, an artificial intelligence model that primarily generates images, such as DALL-E 2 developed by OpenAI, can generate highly realistic images and has the characteristics of being able to imitate various artistic styles. In contrast, an artificial intelligence model called Midjourney has the characteristics of being capable of a high level of artistic expression. However, it may be difficult for users to understand the characteristics of these artificial intelligence models. To facilitate the user's understanding of the characteristics of these artificial intelligence models, in the model list output step (S100), the server (10) transmits sample data regarding the expected results when using the artificial intelligence model selected by the user to the user terminal (20) along with the artificial intelligence model list. The user terminal (20) can output the artificial intelligence model list received from the server (10) and sample data regarding the expected results when using a specific artificial intelligence model. The user can more easily select the artificial intelligence model of his / her choice through the sample data output to the user terminal (20).

[0069] The sample data described above can be stored in a database (30), and the server (10) can transmit the sample data to the user terminal (20) through communication with the database (30).

[0070]

[0071] [Example 5-1] The present invention relates to a method for integrating and using artificial intelligence multiple models. In Example 3-1, the model list output step (S100) outputs sample data on expected results when a plurality of artificial intelligence models selected by a user are combined and used, along with the artificial intelligence model list.

[0072] [Example 5-2] The present invention relates to a method for integrating artificial intelligence multi-models. In Example 5-1, the sample data is pre-stored in a database (30).

[0073] The sample data for the expected results when using the artificial intelligence models described in Examples 4-1 and 4-2 above also applies when a user combines and uses multiple artificial intelligence models. That is, in the model list output step (S100), the server (10) transmits sample data for the expected results when using multiple artificial intelligence models selected by the user in combination to the user terminal (20), and the user terminal (20) can output the received sample data together with the artificial intelligence model list.

[0074] However, since the number of combinations of the selected multiple artificial intelligence models increases as the number of artificial intelligence models increases, there may be a problem in that the transmission speed and the storage space of the user terminal (20) are insufficient to transmit all the combinations of sample data from the server (10) to the user terminal (20) at once. Therefore, when the user preliminarily selects multiple artificial intelligence models on the user terminal (20) in the model list output step (S100), the user terminal (20) transmits information on the multiple artificial intelligence models selected by the user to the server (10), and the server (10) can receive sample data of the combinations of the multiple artificial intelligence models selected by the user from the database (30) and transmit them to the user terminal (20), and the user terminal (20) can output the sample data received from the server (10).

[0075]

[0076] The present invention is not limited to the above-described embodiments, and the scope of application is diverse. It goes without saying that anyone with ordinary skill in the art can make various modifications without departing from the gist of the present invention as claimed in the claims.

Claims

1. A model list output step (S100) that transmits a list of available artificial intelligence models from a server (10) to a user terminal (20) and outputs the list of artificial intelligence models received from the user terminal (20); After the model list output step (S100), when the user selects an artificial intelligence model to be used through the user terminal (20) and inputs data, an optional transmission step (S200) is performed to transmit the artificial intelligence model selected from the user terminal (20) and the input data to the server (10); After the above optional transmission step (S200), a preprocessing step (S300) for preprocessing data received from the server (10); After the above preprocessing step (S300), an execution variable setting step (S400) for setting the execution variables of the artificial intelligence model in the server (10); After the above execution variable setting step (S400), the model execution step (S500) of executing the selected artificial intelligence model based on the data input by the user in the server (10) to obtain a result value; An artificial intelligence multi-model integrated use method including a result output step (S600) for outputting the result value obtained from the server (10) to the user terminal (20) after the model execution step (S500).

2. In paragraph 1, An artificial intelligence multi-model integrated use method, comprising: an accuracy judgment step (S700) for judging the accuracy of the result value in the server (10) after the model execution step (S500) and before the result output step (S600); 3. In paragraph 1, In the above model list output step (S100), the number of artificial intelligence models included in the list of available artificial intelligence models transmitted by the server (10) to the user terminal (20) is at least two or more, The user selects at least two artificial intelligence models to be used through the user terminal (20). An artificial intelligence multi-model integration method comprising an integration method candidate transmission step (S310) for transmitting an artificial intelligence model integration method candidate from the server (10) to the user terminal (20) after the above preprocessing step (S300) and before the above execution variable setting step (S400).

4. In paragraph 1, The above model list output step (S100) is a method for integrating artificial intelligence multi-models, which outputs sample data on expected results when using an artificial intelligence model selected by a user, together with the above artificial intelligence model list.

5. In paragraph 3, The above model list output step (S100) is a method for integrating multiple artificial intelligence models, which outputs sample data on expected results when multiple artificial intelligence models selected by a user are combined and used together with the above artificial intelligence model list.

Images