Chat / generative ai system using generative ai through chat system

Abstract

To provide a chat system in which users who are registered as members of the chat system can interactively exchange, send and receive information such as text, images, URLs, and files, and a chat / generative AI system in which, in order to use the generative AI in the chat system, users who are members of the chat system and the generative AI can interactively respond, and users can input prompts to the generative AI and obtain output.SOLUTION: A method for realizing a multi-headed AI-Bot that can access multiple AIs simultaneously is not to use a conventional Chatbot, but rather to develop a Chatscape window as a function to support the above operations and the handover of content to the Chat system, on the basis of the idea that humans are to directly operate the "Prompt Input and Output" aspect of the generating AI, as desired by the generating AI site.SELECTED DRAWING: Figure 7

Description

[Technical Field]

[0001] This invention relates to a chat system in which users who are registered as members of the chat system can exchange and send / receive information such as text, images, URLs, and files in an interactive format; This relates to a Chat / Generation AI system that enables the users, who are members of the Chat system, to have a dialogue with the Generation AI in order to utilize the Generation AI in the Chat system, and enables the users to input prompts to the Generation AI and obtain output. [Background technology]

[0002] As mentioned above, the Chat-Generative AI system, which uses generative AI through the Chat system, has the following advantages: First, if generative AI could be used in chat systems, which have already become so widespread that they are considered social infrastructure, anyone would be able to use AI easily. It would also be a very convenient method for sharing AI output with other members of the chat system and working in teams, as well as for managing the content produced by generative AI by using the chat system's functions such as recording, saving, playing back, and transferring the history of sent and received information. Secondly, the Chat system can also connect and use AI from multiple developers if necessary. Furthermore, the use of multiple AIs in a chat system allows for cooperative operations such as mutual comparison, collation, selection, and combination of the outputs of multiple AIs, which is an extremely useful method for obtaining high-quality answers from generated AI, such as clearing copyright and intellectual property issues.

[0003] On the other hand, the current Chat generation AI system has two main problems. The first problem is that generative AI services are in a state of rapid evolution and development, and as a result, they are frequently updated in a short period of time. This changes the API required for chatbot access, and each time this happens, a human user's login password is required, making it impossible for chatbots to access the service. Additionally, some generative AI service sites do not allow access via chatbots, i.e., automatic response programs, and may ask "Are you human?" and request a login password using a GUI that is difficult for chatbots to use. Currently, only a few AI generation sites provide APIs that allow users to access and use chatbots (auto-response programs), and most of them use mechanisms to restrict chatbot access and use, as described above. The second problem is that depending on the generative AI, the service may present an answer (output) consisting of multiple options in response to a single question (prompt) from the user, and ask the user to select one or more appropriate options from the multiple options and incorporate them. In such cases, only the user who inputs the prompt can select and decide whether to adopt the multiple options in the output that are output according to the prompt input. With the current scheme, there is no way to confirm the user's selection and decision, making it impossible for a chatbot that responds according to the scenario to handle this. As a result, the chatbot will only receive one option selected by the generating AI site as its answer.

[0004] The purpose of this invention is to maintain the advantages of the chat system while overcoming the problems of the current system and adding new functions to the chat system, thereby developing the chat generation AI system into an essential means of using AI. Building a chatbot for generative AI as a way to interface a chat system with generative AI is a well-known and common method, but All of the problems with the above systems demonstrate the limitations of traditional chatbot functionality. When inputting prompts to the generated AI and obtaining output, if an unexpected response such as selecting an option is required, a conventional chatbot automatic response program cannot handle the situation, and control by the chat system will fail. Furthermore, while anyone can imagine interfacing multiple generative AIs to a chat system to obtain answers to a single question and compare them, the current system does not take into consideration a mechanism for efficiently utilizing multiple AIs simultaneously.

[0005] This invention is a chat / generative AI system that not only overcomes the limitations of chatbots in conventional chat systems, but also maintains control and continuity of the chat system, enabling prompt input and output acquisition from multiple generation AIs simultaneously, thereby significantly expanding the functionality of the chat system.

[0006] As mentioned above, when conventional chatbot auto-response programs are unable to handle the "prompt input and output" of the generated AI, the system of this invention provides a "window (we will call this window Chatscape) that supports the user's operation of "prompt input and output" for the generated AI and performs interface processing to the chat system, including converting the acquired output into chat normalized data," enabling the generated AI to operate in a manner that is compatible with the generated AI. We will then build a Chatscape Chatbot that will allow users of the chat system to access and use Chatscape. We will call this Chatscape Chatbot the Chatscape Bot. The greatest effect of this invention is that in the Chatscape window, the user can sequentially switch between multiple generated AIs that interface with the chat system, thereby realizing the operation of obtaining the output of multiple generated AIs for one prompt. Through user operation and settings, it becomes possible to turn the Chatscape Bot into a multi-headed AI-Bot that combines multiple AIs and behaves like a single AI-Bot, which is unprecedented, and not only overcomes the problems of conventional chatbot technology but also greatly expands the functionality of the chat system.

[0007] Next, we will explain the Chatscape window of this invention based on an actual operating screen. In [Figure 1], Chatscape places the Chat system's in-app browser window [Figure 1-▲1▼] on the left side of the screen for the generated AI's "Prompt Input and Output," which the user must directly operate, and places a relay window [Figure 1-▲2▼] on the right side, which exchanges content between the browser on the left and the Chat system. In a working example, the user first sends a prompt to the generated AI to the Chatscape Chatbot, i.e., Chatscape-Bot. When Chatscape receives the prompt, it translates it into English, and when the user is ready to directly operate the Chatscape window to perform "Prompt Input and Output," Informing the user of the acknowledgement; The English translation of the Prompt is posted below. An acknowledgement balloon with the Chatscpape start button is sent and displayed in the chat room on the user's terminal. [Figure 2] In another example of actual operation, the AI ​​used supported API, so it was possible to operate in two modes: API mode, which uses the AI ​​used, and Chatscape mode, as described above. Therefore, we created an acknowledgement balloon that supports both modes. [Figure 3] The operation and the acknowledgement balloon that support only the Chatscape mode and the operation and the acknowledgement balloon that support both the API and Chatscape modes are as follows: The latter has the same functionality as the former acknowledge balloon, except that it allows you to execute API-compliant "Prompt input and output" by pressing an API button. In the Chatscape window that was started by pressing the Chatscape start button located in the acknowledgement balloon, By entering the URL you want to open in the address bar at the top of the browser window, the specified URL will open. Users can also manually tap the generated AI icon on the browser screen to change the URL in the address bar to the corresponding generated AI. They can also select a URL from the drop-down menu and set it in the same way. When starting operation by inheriting the prompt issued by the user from the chat system and the URL of the target generated AI, it is also possible to set the address URL of the generated AI in question in the address bar at the top of the browser [Figure 1-▲3▼]. In that case, when operation starts, the generated AI screen opened with this URL will wait for the user to enter a prompt. The relay window on the right [Figure 1-▲2▼] displays and stores the prompt for the generated AI inherited from the chat system, and waits for user input just like the browser. The prompts displayed and stored in the relay window are also stored in the copy and paste board, so the user can simply paste them into the prompt input box on the browser screen to transcribe and input them, and the "Prompt input and output" action for the generated AI is performed.

[0008] [Figure 4] is a Chatscape screen showing the output. When the generated AI receives the prompt input information, it will output the corresponding output on the browser screen [Figure 4-▲1▼]. The user can simply press the copy button on the browser to move the output to the relay window [Figure 4-▲2▼], where the Chat normalized data is processed and control is passed to the Chat system. In [Figure 4] and [Figure 5], if the Output consists of multiple Options, select each option one by one, press the copy button on the browser, and move it to the Relay Window [Figure 4-▲2▼]. The Relay Window's bottom-up array loading function will load each Output Option into the Relay Window's bottom-up array mechanism list [Figure 4-▲2▼] and [Figure 5] one by one. In addition, various reference information, images, and other AI outputs displayed on the output screen of the generated AI can be imported as needed by simply pressing the copy button. Furthermore, users can easily add their own comments and alternatives to the output to the list of this bottom-up arrangement mechanism [Figure 5] by opening the URL in a browser and displaying it on the screen. The bottom-up array mechanism list [Figure 5] information in the relay window [Figure 5-▲2▼] is stored in the payload as chat sending and receiving information for the chat system along with the original prompt information, site information for the generating AI, and channel information for the chat system for chat normalization data processing, and is sent to the chat system's control server, recorded in the chat system's history database (i.e., chat normalization data processing), and sent to the chat room on the user's terminal for display. In the chat room [Figure 6], the output of the generated AI is displayed as an array balloon in the same format as the bottom-up array in the relay window. In this way, the series of "Prompt Input and Output" actions for one generated AI that was initially set up in the Chatscape browser are configured so that the same actions can be performed for a different generated AI by switching the Chatscape browser.

[0009] In the previous explanation of the Chatscape window, we explained how Chatscape handles AI "prompt input and output" behavior, which previous Chatbot technology was unable to handle. This invention not only overcomes the weaknesses of conventional technology, but also displays the icons of multiple generation AIs available in the chat system on the browser screen in the Chatscape window, allowing users to select the generation AI to operate by tapping on one of the generation AI icons, making it possible to efficiently and quickly obtain the output of multiple generation AIs for a single prompt task. In this way, depending on the user's operations and settings, Chatscape Bot appears from the Chat system's perspective as a single Bot with multiple AI brains, i.e., a multi-headed AI-Bot, operating like the Chat room in Figure 5-▲2▼. Next, based on an actual example of incorporating Chatscape, This diagram explains the process from when the user issues the "prompt" message to when the resulting output is finally displayed on the user's terminal. (Total Process first half schematic diagram) The first half of the overall process is a diagram of the chat system terminal user sending a prompt to the Chatscape-bot for the AI ​​to generate. The Chatscape-Bot receives the prompt and displays it in the Chatscape window. This shows the process flow from the Chat system to the control that supports the user's "Prompt Input and Output" operations and behaviors. [Figure 8] Schematic diagram of the second half of the overall processing (Total Process later half schematic diagram) The overall processing second half schematic diagram is In the Chatscape window, the user completes the "Prompt Input and Output" operation for the generated AI and transfers the acquired Output to the Chat system. This shows the process from when the Chatscape-Bot delivers the Output to the user's chat room until it is displayed. [Brief explanation of the drawings]

[0010] [Figure 1] An example of a screen where the Chat system's in-app browser window and the relay window are placed in the Chatscape window. [Figure 2] An example of a screen where an acknowledgement balloon with a Chatscpape start button is sent and displayed in the chat room on the user's device. [Figure 3] An example screen showing two modes in a chat room on a user's device: API mode, which uses generated AI, and Chatscape mode, along with an acknowledgement balloon that supports both modes. [Figure 4] Chatscape screenshot showing output [Figure 5] Relay window bottom-up arrangement mechanism list in Chatscape window and example of bottom-up arrangement balloon display in terminal chat system [Figure 6] In a chat room, the output of the generation AI is displayed as an array balloon in the same format as the bottom-up array in the relay window. [Figure 7] Overall processing first half schematic diagram [Figure 8] Schematic diagram of the second half of the overall process

Claims

1. Users who are registered as members of the chat system can exchange information interactively. In a Chat system that exchanges and sends and receives messages, In order to enable the user to use the generated AI, the app has an in-app browser that can be operated directly and manually by the user and that can browse any designated site, and a relay window that is displayed in the browser and passes selected content to the Chat system, and a Chatscape window that takes over control of "Prompt Input and Output" including prompt sentences issued by the user for the generated AI from the Chat system, When the Chatscape receives a prompt for the generated AI from the user, an acknowledgement balloon that notifies the user that the Chatscape is ready to be operated by the user and includes a start button for starting the operation process of the Chatscape; The user issues a prompt to the generation AI, A Chat / generation AI system in which, when the Chatscape receives a prompt or other message from the user, it issues an acknowledgement balloon, allowing the user to directly operate the Chatscape to obtain the output of the generation AI and to hand over the output to the Chat system for display.

2. Users who are registered as members of the chat system can exchange information interactively. In a Chat system that exchanges and sends and receives messages, In order to enable the user to use any plurality of generated AIs from different developers, the app has an in-app browser that can be directly and manually operated by the user and can browse any designated site, and a relay window that is displayed in the browser and passes selected content to the Chat system, A Chatscape window that takes over control of "Prompt Input and Output" including the generated AI destination unspecified prompt sentence issued by the user from the Chat system, When the Chatscape receives a prompt from the user that does not specify the AI ​​destination, an acknowledgement balloon that notifies the user that the Chatscape is ready to be operated by the user and includes a start button for starting the operation process of the Chatscape; The user issues a prompt to the AI ​​generator without specifying a destination. When the Chatscape receives the prompt text etc. from the user, it issues the acknowledgement balloon, When the in-app browser of the Chatscape is in a state where a generated AI has not been set, the user directly operates the in-app browser to determine and set a generated AI that has not been set, input the prompt text to the generated AI, and obtain output. The user also operates the in-app browser again to switch the previously set generated AI and set a new, different generated AI, input the prompt text to the different generated AI, and obtain output. The user repeats the operation to obtain multiple outputs from multiple generation AIs, such as obtaining the output of another generation AI. A chat generation AI system that allows data to be handed over and displayed on the chat system.

3. Users who are registered as members of the chat system can exchange information interactively. In a Chat system that exchanges and sends and receives messages, In order to enable the user to use the generated AI, the app has an in-app browser that can be directly and manually operated by the user and can browse any designated site, and a relay window that is displayed in the browser and passes selected content to the Chat system, Control of "Prompt Input and Output" including prompt sentences issued by the user to the generated AI A Chatscape window that takes over from the Chat system and operates, When the Chatscape receives a prompt for the generated AI from the user, An acknowledgement balloon is provided which notifies the user that the user is ready to operate the Chatscape and has two buttons: a Chatscape button which the user uses to select a transition to the Chatscape operation process, or an API button which the user uses to select an operation process conforming to the API defined by the generation AI; The user issues a prompt to the generated AI, and when the Chatscape receives the prompt from the user, it issues the acknowledgement balloon. When the user presses the Chatscape button, the user can directly operate the Chatscape to obtain the output of the generated AI. When the user presses the API button, the API-compliant automatic response program sends a prompt to the generated AI to obtain the output. A Chat generation AI system that makes it possible to hand over and display the output obtained by any of the above methods to the Chat system.

Images