AI-powered intelligent systems with multiple output data that can reduce erroneous selections

Abstract

This invention discloses an AI intelligent system with multiple output data that can reduce erroneous selections, comprising: a registration module for inputting a device type or payment type registered by a user; an initialization module connected to the registration module, which configures multiple different artificial intelligence models according to the device type or payment type, wherein each artificial intelligence model is based on a large language model; an input module connected to the initialization module for inputting a user request; a processing module connected to the input module, which assigns the user request to the multiple different artificial intelligence models; a result generation module connected to the processing module, which generates multiple results from the user request using the multiple different artificial intelligence models; and a result display module connected to the result generation module, which displays the multiple results.

Description

Technical Field

[0001] This invention relates to an AI intelligent system, and more particularly to an AI intelligent system with multiple output data that can reduce erroneous selections. Background Technology

[0002] Artificial intelligence systems (AI systems) are systems that simulate human intelligence through computers, aiming to perform tasks requiring human thought, such as learning, reasoning, understanding, decision-making, language processing, perception, and interaction. These systems typically operate based on technologies such as machine learning, deep learning, natural language processing (NLP), and computer vision, and can accomplish various tasks according to different application scenarios.

[0003] For example, Taiwan Patent No. I823785, entitled "A Method for Ranking the Best Answers to a Goodwill Question Using Generative Artificial Intelligence," includes providing a client's own system; the client's own system receiving a first query; the client's own system obtaining the answer to the first query from a generative artificial intelligence server, wherein the generative artificial intelligence server is connected to the client's own system via the Internet; transmitting the answer obtained from the generative artificial intelligence server back to the client's own system, and using the answer as a second query to search within the client's own system; the client's own system searching for documents related to the second query; and performing a relevance ranking on the related documents.

[0004] However, please refer to case number I823785. Figure 3 Current technology relies on a single AI model within a generative AI server to generate a single result. This is because the Chat Generative Pre-trained Transformer (ChatGPT) is a conversational AI model developed based on a Large Language Model (LLM). LLMs have the following characteristics: 1. Language simulation rather than true understanding: LLMs primarily rely on statistical patterns and data learning, generating text that conforms to language rules but lacking a true understanding of the intent or logic behind the text. 2. Lack of common-sense reasoning: While LLMs can provide reasonable answers in most situations, their understanding of common sense or general knowledge remains flawed, sometimes leading to logical errors or misunderstandings. 3. Bias in training data: LLMs are trained on large amounts of text data, which may contain various biases from human society (such as gender, race, and culture). Therefore, LLMs may generate biased, discriminatory, or inappropriate content, making it prone to errors when generated by a single AI model.

[0005] Therefore, how to eliminate the above-mentioned deficiencies is the technical difficulty that this invention aims to solve. Summary of the Invention

[0006] In view of the above, the purpose of this invention is to solve and improve the problems and deficiencies of the prior art by providing an AI intelligent system with multiple output data that can reduce erroneous selection.

[0007] To achieve the above objectives, the present invention provides an AI intelligent system with multiple output data that can reduce erroneous selections. The system includes: a registration module for inputting a device type or payment type registered by a user; an initialization module connected to the registration module, which configures multiple different artificial intelligence models based on the device type or payment type, each of which is based on a Large Language Model (LLM); an input module connected to the initialization module for inputting a user request; a processing module connected to the input module, which assigns the user request to the multiple different artificial intelligence models; a result generation module connected to the processing module, which generates multiple results from the user request using the multiple different artificial intelligence models; and a result display module connected to the result generation module, which displays the multiple results.

[0008] It also includes a storage module that is connected to the results display module and stores the user request and the multiple results.

[0009] These multiple AI models include Chat Generative Pre-trained Transformer (ChatGPT), Gemini, and Shunfei.

[0010] Specifically, the processing module selects at least two different artificial intelligence models from the model library through an API Gateway based on the user's request.

[0011] The user request is based on voice input, which is provided by a microphone.

[0012] The user request is based on a text input provided by a prompt library.

[0013] This device type includes OAI-Mouse (Open Accelerator Infrastructure Mouse), OAI-Keyboard (Open Accelerator Infrastructure Keyboard), OAI-Pen (Open Accelerator Infrastructure Pen), and OAI-Wearable Device (Open Accelerator Infrastructure Wearable).

[0014] This payment type includes cash transfer, credit card payment, or electronic payment.

[0015] Existing technologies generate a single result from a single artificial intelligence model, which is prone to producing erroneous results. In contrast, this invention generates multiple results for the user's request using multiple artificial intelligence models in the model library, allowing the user to select the most common result from these multiple results. By using the majority result, the probability of error is reduced, thus effectively reducing the chance of incorrect selection. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the mold composition of the present invention.

[0017] Figure 2 This is a hardware connection configuration diagram for the present invention.

[0018] Figure 3 This is a flowchart of the operation steps of the present invention.

[0019] Figure 4 This is a schematic diagram of the display interface in this invention.

[0020] Figure 5 This is a flowchart illustrating the operation of the present invention.

[0021] Figure label explanations: 1-Registration module; 2-Initialization module; 3-Input module; 4-Processing module; 5-Result generation module; 6-Result display module; 7-Storage module; 110-AI IO device; 120-Backend device; 130-Frontend device; 200-Display interface; S1, S2, S3, S4, S5-Steps; A1-Users can configure different artificial intelligence models; A2-Users register, log in, and start the system; A3-Perform system initialization; A4-Input user request; A5-Users can input voice through the microphone of the OAI mouse, or use the system's built-in Prompt library to input the user request; A6-After the user submits the user request, the system will send the user request to the backend device for processing through the API gateway. Based on the load balancing strategy, the user request is distributed to the appropriate device. Artificial intelligence model processing; A7 - User request assigned to the first artificial intelligence model; A8 - User request assigned to the second artificial intelligence model; A9 - User request assigned to the third artificial intelligence model; A10 - User request assigned to the fourth artificial intelligence model; A11 - First result generation; A12 - Second result generation; A13 - Third result generation; A14 - Fourth result generation; A15 - The system generates corresponding results based on user input and configuration, and transmits the results to the front-end device for display; A16 - Result display; A17 - Result adjustment and summary; A18 - Users can evaluate or score the results, and can also select results for further summary and comparison. Detailed Implementation

[0022] To facilitate a concise understanding of the other features, advantages, and effects of the present invention, the present invention will be described in detail below with reference to the accompanying drawings:

[0023] Please see Figure 1 As shown, the present invention provides an AI intelligent system for multiple output data that can reduce erroneous selection, comprising:

[0024] A registration module 1 is provided for users to input either a device type or a payment type for registration. The device type includes OAI-Mouse (Open Accelerator Infrastructure Mouse), OAI-Keyboard (Open Accelerator Infrastructure Keyboard), OAI-Pen (Open Accelerator Infrastructure Pen), and OAI-Wearable (Open Accelerator Infrastructure Wearable), but is not limited to these in practice. Open Accelerator Infrastructure (OAI) is an open-source hardware organization that promotes the standardization of AI hardware. An OAI-Mouse refers to a dedicated mouse designed according to the standards published by Open Accelerator Infrastructure. In addition to traditional clicking and scrolling, it includes artificial intelligence software, sensors, and a wireless receiver, supporting gesture control or other innovative functions. It also supports wired and wireless connections and can collaborate with various devices and platforms, such as one-click access to ChatGPT. The payment type includes cash transfer, credit card payment, or electronic payment, but is not limited to these in practice.

[0025] An initialization module 2 is connected to the registration module 1. This initialization module 2 configures multiple different artificial intelligence models based on the device type or payment type, and each artificial intelligence model is based on a Large Language Model (LLM). These multiple artificial intelligence models include Chat Generative Pre-trained Transformer (ChatGPT), Gemini, and Shunfei, etc., which are only examples and do not limit the application scope of this invention. ChatGPT is a conversational artificial intelligence model developed by OpenAI based on a large language model. Gemini is an artificial intelligence model developed by Google based on a large language model.

[0026] An input module 3, connected to the initialization module 2, is provided for inputting a user request. This user request can be based on either voice input or text input. Voice input can be provided by a microphone, while text input can be provided by a prompt library. A prompt can be understood as providing the AI ​​model with a text, a sentence, or a question, guiding the AI ​​model to respond to the question in a more accurate and in-depth manner.

[0027] A processing module 4, connected to the input module 3, distributes the user request to multiple different AI models. Specifically, the processing module 4 distributes the user request to the multiple different AI models via an API gateway. This API gateway is an intermediary layer used to process and manage all Application Programming Interface (API) requests in the application, enabling traffic distribution and forwarding the user request to the multiple different AI models as needed, achieving load balancing.

[0028] A result generation module 5 is connected to the processing module 4. The result generation module 5 generates multiple results from the user request using the multiple different artificial intelligence models.

[0029] A result display module 6 is connected to the result generation module 5, and the result display module 6 displays the multiple results.

[0030] It further includes a storage module 7, which is connected to the result display module 6, and the storage module 7 stores the user request and the multiple results.

[0031] Please see Figure 2As shown, this invention is applied in a network, and its hardware includes an AI IO device 110, a back-end device 120, and a front-end device 130, which work together in cooperation through the back-end device 120, the front-end device 130, and back-end processing. The AI ​​IO device 110 refers to an input / output (IO) device related to artificial intelligence (AI) to support AI data collection, processing, and result output functions. The AI ​​IO device 110 may include an OAI mouse, an OAI keyboard, an OAI pen, or an OAI wearable device. The back-end device 120 may include servers, storage devices, routers and switches, load balancers, and virtualization platforms, and the virtualization platform may be a cloud platform with multiple artificial intelligence models, including ChatGPT, Gemini, and Shunfei. The backend device 120 is responsible for processing requests from the frontend device 130, performing data storage, logical processing, authentication, calculations, etc., and returning the results to the frontend device 130. The frontend device 130 refers to the device or interface that directly interacts with the user, used to display application content, provide a user interface, and capture user input. The frontend device 130 may include desktop devices and mobile devices. Desktop devices may include, for example, desktop computers, laptops, and monitors. Mobile devices may include, for example, smartphones and tablets. The backend processing refers to all data calculations, logical processing, and data interaction operations responsible for processing and managing requests sent from the frontend device 130. The backend processing is closely related to the backend device 120, which is typically the actual hardware or infrastructure executing the backend processing. Simply put, the backend processing refers to the process of performing data calculations, logical processing, and data interaction operations, while the backend device 120 is the device that executes these operations. The backend processing includes receiving a request, processing logic, querying data, generating a response, and returning a result to the frontend. The "receive request" refers to the request being sent to the cloud platform's server when a user sends it (e.g., clicking a button or filling out a form). The "processing logic" refers to the server's actions upon receiving the user request, based on pre-defined application logic, such as querying the database, calculating results, or providing other services. The "data query" refers to the backend device 120 sending a query to the database if necessary and processing the query results.The generated response is constructed by the backend device 120 based on the result of request processing. The return result to the frontend means that the backend device 120 sends the response back to the frontend device 130, and the frontend device 130 displays the result.

[0032] Please see Figure 1 , Figure 2 , Figure 3 and Figure 5 As shown, Figure 5 for Figure 3 The process description is used to assist in explaining steps S1 to S5 of the flowchart, so the same content will not be repeated. The steps of this invention include:

[0033] Step S1 involves configuring multiple AI models based on the user's device type or payment type, with the user able to configure different AI models A1 independently. Step S1 involves the collaborative work of front-end device 130 and back-end device 120. Front-end device 130, through registration module 1, displays the registration and login interface and collects user input data (username, password, email, device type, or payment type). After the user completes input on the front-end device 130's interface, the front-end device 130 sends the registered device type or payment type to the back-end device 120. The front-end device 130 then sends the required registration or login data (username, password, email, device type, or payment type) to the back-end device 120. The back-end device 120 handles the processing logic and data storage. Upon receiving a request from front-end device 130, the back-end device 120 processes the registration or login logic.

[0034] For example, during registration, backend device 120 checks if the username already exists and stores the new user's data (such as an encrypted password) in the database. During login, backend device 120 verifies the correctness of the username and password and typically generates an authentication token (such as a JWT) to return to frontend device 130. Furthermore, not all AI models can accept images as input. For example, ChatGPT can process text and images and answer questions based on image content. Conversely, not all AI models can accept voice as input. For instance, Google Assistant can interact with users via voice commands, answer questions, set reminders, etc. For example, if a user's device type is an OAI-mouse (with a built-in microphone) and payment type is credit card, the user can configure ChatGPT and Google Assistant themselves. At this point, ChatGPT and Google Assistant are not actually configured for the user; only the user's device type or payment type is linked to ChatGPT and Google Assistant. The actual configuration will only be implemented after login.

[0035] Step S2 involves user registration and login, system startup A2, system initialization A3, and input of user requests A4. Step S2 involves the collaborative work of front-end device 130 and back-end device 120. Front-end device 130 is responsible for initialization, including displaying the login interface and homepage. When a user logs in, front-end device 130 sends an initialization request to back-end device 120 through initialization module 2. After receiving the initialization request from front-end device 130, back-end device 120 configures multiple different AI models based on the user's registered device type or payment type. Only then are ChatGPT and Google Assistant actually configured for the user. The user request is initiated by front-end device 130 and processed by back-end device 120. Users can input via voice through the microphone of the OAI-mouse or use the system's built-in Prompt library to input user requests A5. Front-end device 130 uses input module 3 to input user requests, captures the user requests, and sends them to back-end device 120 for further processing.

[0036] In step S3, after the user submits a user request, the system sends the user request to the backend device 120 for processing through the API gateway. Based on the load balancing strategy, the user request is assigned to the appropriate artificial intelligence model A6 for processing. Specifically, in step S3, after the user inputs and submits the user request, the system forwards the user request to the backend device 120 for processing through the API gateway via the processing module 4. The API gateway is responsible for processing and routing the user request (routing is the process by which the user request determines how to be sent to multiple artificial intelligence models). The API gateway forwards the user request to the backend device 120 for corresponding services or processing based on the type of user request and the device type. For example, the user request may be assigned to the first artificial intelligence model A7, which could be, for example, ChatGPT; the user request may be assigned to the second artificial intelligence model A8, which could be, for example, Gemini; the user request may be assigned to the third artificial intelligence model A9, which could be, for example, Shunfei; and the user request may be assigned to the fourth artificial intelligence model A10, which could be, for example, Google Assistant. In short, the same user request is assigned to four different artificial intelligence models.

[0037] In step S4, the AI ​​model generates corresponding results based on the user's input request and configuration, and transmits the display results to the front-end device 130. The front-end device 130 then displays the results. Step S4 involves the back-end device 120 processing the user request's specific business logic, such as multiple AI models verifying database queries and performing operations. The result generation module 5 generates multiple results, which are then transmitted to the front-end device 130 for display. The front-end device 130 receives these multiple results from the back-end device 120 and displays them to the user through the result display module 6. For example, the user request might generate the first result (A11) using ChatGPT; the second result (A12) using Gemini; the third result (A13) using Shunfei; and the fourth result (A14) using Google Assistant. In short, the same user request generates results using four different AI models. Afterward, the system generates corresponding results based on the user's input and configuration, transmits the display results (A15) to the front-end device, and the front-end device 13 displays the results (A16).

[0038] In step S5, the user sees the results processed by the artificial intelligence model on the front-end device 130. The user can adjust and summarize the results (A17). The user can evaluate or score the results, or select results for further summary and comparison (A18).

[0039] Please see Figure 2 and Figure 4 As shown, Figure 4 The display interface 200 of the front-end device 130 can accept voice or text input from the AI ​​IO device 110 to connect to the back-end device 120 and access various artificial intelligence models to obtain the following user applications:

[0040]

[0041] Multi-AI module chat refers to a dialogue system that uses multiple AI modules to interact or collaborate. Each AI module typically has a different area of ​​expertise or capability and can work independently or collaboratively to complete certain tasks or provide information.

[0042] Voice memos are an application based on speech recognition and natural language processing (NLP) technologies that can convert speech into text and process, store, manage, and search it.

[0043] Voice recording is the process of saving voice or sound.

[0044] Summarizing and organizing involves simplifying and refining the key points of a piece of information, an article, a speech, or a discussion.

[0045] Mind mapping is a visualization tool that uses images to organize and display information.

[0046] AI-powered image cropping is a tool or function that uses artificial intelligence technology to automatically or semi-automatically crop images. Through deep learning algorithms and computer vision technology, it can identify different elements or regions in an image and intelligently crop or remove unwanted parts to achieve more accurate or more demanding results.

[0047] The jump server allows direct login to the existing AI artificial intelligence module.

[0048] This invention has the following characteristics:

[0049] I. Multi-Model Integration: Supports multiple different device types, using standardized AI IO devices 110 to ensure the standardization of input / output of multiple different artificial intelligence models.

[0050] II. Dynamic Model Selection: Multiple different artificial intelligence models can be assigned to different users, and users can configure them themselves.

[0051] 3. Extensibility: With the integration of more artificial intelligence models, it can have good extensibility.

[0052] Therefore, existing technologies generate a single result from a single artificial intelligence model, which is prone to producing erroneous results. In contrast, the present invention generates multiple results for the user's request using multiple artificial intelligence models, allowing the user to select the most common result from these multiple results. By using the majority result, the probability of error is reduced, thereby effectively reducing the chance of incorrect selection.

[0053] The above discussion is merely a preferred embodiment of the present invention and is not intended to limit the scope of the present invention. Therefore, any equivalent modifications, combinations, substitutions, or modifications made without departing from the spirit and scope of the present invention should be covered within the protection scope of the present invention.

Claims

1. An AI intelligent system with multiple output data that can reduce erroneous selection, characterized in that, Include: A registration module, which allows a user to input either a device type or a payment type for registration; An initialization module is connected to the registration module. The initialization module configures multiple different artificial intelligence models according to the device type or the payment type, and each artificial intelligence model is based on a large language model. An input module is connected to the initialization module and is used to input a user request. A processing module connected to the input module, which assigns the user request to the multiple different artificial intelligence models; A result generation module, which is connected to the processing module, generates multiple results from the user request using the multiple different artificial intelligence models; as well as A result display module is provided, which is connected to the result generation module and displays the multiple results.

2. The AI ​​intelligent system for reducing erroneous selection of multi-output data as described in claim 1, characterized in that, It also includes a storage module connected to the results display module, which stores the user request and the multiple results.

3. The AI ​​intelligent system for reducing erroneous selection of multi-output data as described in claim 1, characterized in that, These multiple AI models include chat-generating pre-trained transformation models, Gemini, and Shunfei.

4. The AI ​​intelligent system for reducing erroneous selection of multi-output data as described in claim 1, characterized in that, The processing module selects at least two different artificial intelligence models from the model library through an API gateway based on the user's request.

5. The AI ​​intelligent system for reducing erroneous selection of multi-output data as described in claim 1, characterized in that, The user request is based on voice input, which is provided by a microphone.

6. The AI ​​intelligent system for reducing erroneous selection of multi-output data as described in claim 1, characterized in that, The user request is based on a text input provided by a prompt dictionary.

7. The AI ​​intelligent system for reducing erroneous selection of multi-output data as described in claim 1, characterized in that, This device type includes OAI-mouse, OAI-keyboard, OAI-pen, and OAI-wearable devices.

8. The AI ​​intelligent system for reducing erroneous selection of multi-output data as described in claim 1, characterized in that, This payment type includes cash transfer, credit card payment, or electronic payment.

Images