Prompt sharing method and apparatus

Abstract

Disclosed are a prompt sharing method and apparatus, aiming to realize efficient sharing and personalized processing of generalized prompts (comprising multi-modal information such as texts, pictures, voices, and videos). By means of interaction between a sharer and a recipient, combined with a specified artificial intelligence large model, prompts are processed automatically and a result is generated. The present invention can realize rapid sharing and precise personalization of prompts and flexible use of results, helping to improve the application efficiency of artificial intelligence services, and promoting the popularization of large model technology. Prompt sharing will be a very valuable work in the field of artificial intelligence.

Description

Method and device for sharing prompt words Technical Field

[0001] The present invention belongs to the field of artificial intelligence technology, and in particular relates to a method and device for sharing prompt words, which are used to improve the application efficiency and convenience of multimodal artificial intelligence large models. Background Art

[0002] Large AI models use prompts as input, which consist of a purpose and reference information. For multimodal AI models, prompts broadly encompass multimodal information types such as images, audio, and video. Due to the nature of large models, varying prompt quality can lead to significant differences in output, even spawning specialized roles dedicated to writing prompts.

[0003] After a large number of high-quality prompt words are generated, how to quickly and conveniently share these prompt words so that more people can enjoy the better services of the large model will be a very valuable task in the field of artificial intelligence. Summary of the Invention

[0004] The purpose of the present invention is to provide a universal and convenient method and device for sharing artificial intelligence prompt words.

[0005] This method can quickly transmit high-quality prompts from the sharer to the recipient, and generates results through a large artificial intelligence model. Its advantages include simplifying the prompt word transmission process, enhancing the user experience, supporting multi-party collaboration, and optimizing the personalized processing of prompt words.

[0006] To implement the above functionality, a "sharer A" (hereinafter referred to as A), a "recipient B" (hereinafter referred to as B), and a high-quality "generalized prompt word set P" (hereinafter referred to as P or prompt word P) are required. P consists of one or more prompt words. For example, P can be a single text prompt word, a text prompt word and an image prompt word, or multiple prompt words arranged in a conversational sequence. A shares prompt word P with B using methods including but not limited to QR codes, links, and push notifications. Prompt word P is received by a designated "AI big model M" (hereinafter referred to as M or model M), which processes it sequentially and produces the model's output. M is specified by B or agreed upon in advance.

[0007] This method allows the recipient B to automatically use the high-quality prompt word P, thereby avoiding B manually copying all or part of the prompt word P. In order to achieve sufficient ease of use and convenience, the sharing process should not have an explicit download or installation process.

[0008] The specific implementation steps are:

[0009] S1 "Sharer A" shares "generalized prompt word set P" with "receiver B". "Generalized prompt word set P" consists of one or more prompt words;

[0010] S2 "generalized prompt word set P" will be received by the designated "artificial intelligence model M";

[0011] The "artificial intelligence large model M" specified by S3 processes the prompt word P and gives the model result.

[0012] S4 sharing methods include but are not limited to QR codes, links, and push notifications.

[0013] The above steps can simply realize the intention of the present invention, but in order to be more universal, in addition to the general prompt word P, the shared information also requires some "prompt word description information D" (hereinafter referred to as D). P is used for large models, and D is used for non-large models. The content of D includes but is not limited to the specified model, author, price, evaluation, judgment conditions and control instructions related to the prompt word. The judgment condition refers to the operating conditions for judging certain prompt words. For example, a prompt word P related to a physical examination report. If B is a male, the prompt words related to gynecological items in P should not be used. Control instructions refer to instructions for interacting in an environment related to B. For example, the control instructions in D can require the device used by B to take a photo and use the photo as a prompt word. The implementation method is as follows:

[0014] S5 The information shared by "sharer A" to "recipient B" will also include "prompt word description information D". The content of "prompt word description information D" includes but is not limited to the specified model, author, price, evaluation, prompt word related judgment conditions and control instructions.

[0015] To maximize the applicability of this method, the prompt words in P can also be personalized. Personalization can be categorized as pre-sharing or post-sharing. Pre-sharing personalization refers to A personalizing P based on various information. Post-sharing personalization refers to B personalizing P based on various information after receiving the shared document and before M processes it. This information includes, but is not limited to, A's information, B's information, and environmental information. Personalization modification refers to modifying, deleting, or adding prompt words in P. For example, if A shares P with B and it contains A's personal ID number, they can "modify" the relevant prompt words in "Personalize before Sharing" to replace the ID number. Another example is if A shares P with B, which contains a set of prompt words for medical examination report analysis, including both male and female specialty analyses. B wants to analyze his or her own medical examination report. If B is male, the prompt words related to the female specialty in P need to be deleted. Deleting prompt words can, in some cases, have the same effect as the "judgment condition" mentioned above.

[0016] Adding prompt words is a bit complicated. The prompt words need to be combined with the large model (which can be a model other than the model M specified above). By asking the large model about its next plan, new prompt words are formed and added to P.

[0017] The implementation method of personalizing P is as follows:

[0018] S6 personalizes the "prompt word set P" into pre-sharing personalization and post-sharing personalization.

[0019] S7 Personalized modification of the "prompt word set P", including modifying a prompt word and deleting a prompt word.

[0020] S8 Personalized modification of the "prompt word set P", including adding a prompt word.

[0021] The results generated by M are typically used by B, but access to the generated results can also be specified to facilitate multi-person collaboration. For example, the results generated by M can be returned to A or any third party for use. This process usually requires B's consent.

[0022] S9 can specify the user of the results given by the "artificial intelligence big model M", and the user includes but is not limited to "sharer A", "sharer B" or a third party.

[0023] The device for applying artificial intelligence involved in the present invention includes the following modules:

[0024] The D1 storage module is responsible for storing the "prompt word set PA" of "sharer A", the "prompt word set PB" received by "sharer B", the address of the designated "artificial intelligence model M", the results returned after processing by the model, and the intermediate addresses generated during sharing;

[0025] The D2 forwarding module forwards the PA in D1 to the PB in D1 for storage, forwards the PB to the M in D1 to request large model processing, and retrieves the results processed by the M in D1 for storage.

[0026] Optional modules will also include:

[0027] D1 will also include the "prompt word description information D" corresponding to PA and PB.

[0028] The D3 processing module modifies, deletes and adds PA and PB in D1 as required. Beneficial effects

[0029] The method and device of the present invention have the following beneficial effects:

[0030] 1. Ability to quickly share the results of work corresponding to high-quality tips;

[0031] 2. It is beneficial to the rapid implementation of large-scale model applications;

[0032] 3. Contribute to the formation and prosperity of the large model market;

[0033] 4. Fully reduce the cost of implementing artificial intelligence. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 shows the basic logic diagram of the method. Best Mode for Carrying Out the Invention

[0035] To achieve efficient, personalized, and secure sharing of prompt words, the present invention proposes the following optimal implementation methods:

[0036] 1. Multimodal support for prompt words

[0037] The prompt word set P of the present invention is not limited to text prompt words, but also supports multi-modal input such as pictures, voice and video. For example:

[0038] •Text prompts: For example, in the e-commerce field, the product recommendation prompt is “Recommend a smartphone under 3,000 yuan.”

[0039] •Image prompt words: For example, in the medical field, upload a picture of a physical examination report for health analysis.

[0040] • Voice prompts: For example, in the legal consulting field, voice prompts can be used to input “preliminary opinions on contract disputes”.

[0041] •Video prompt words: For example, in the education field, upload teaching video clips and generate learning reports based on the video content.

[0042] 2. Seamless interaction of prompt word sharing

[0043] Through QR codes, links, or push notifications, sharer A can quickly share prompt word P with recipient B without manually copying or installing any additional software. The optimal process is as follows:

[0044] •Sharer A generates a prompt word set P and adds prompt word description information D.

[0045] •Sharer A chooses the sharing method, such as generating a QR code and sending it to recipient B via social media.

[0046] •After the receiver B scans the QR code or clicks the link, the prompt word P is directly loaded into the designated large model M and the task is automatically executed.

[0047] 3. Personalized dynamic adjustment of prompt words

[0048] The best implementation includes both pre-sharing and post-sharing personalization:

[0049] • Personalization before sharing: Sharer A can adjust the prompt words based on privacy or environmental needs before sharing. For example, remove the part containing personal sensitive information.

[0050] • Personalization after sharing: Recipient B can further modify the prompt words according to actual needs before processing by the large model M. For example, adding auxiliary information or restrictions for specific business scenarios.

[0051] 4. Multi-party specification of large model generation results

[0052] To support collaboration, the present invention allows for the generation of specified users of results. For example:

[0053] • Shareholder A designates the results for his own use in order to confirm their accuracy;

[0054] •Receiver B uses the results directly to complete the task;

[0055] •The results can be designated for use by a third party C, such as an expense report generated by financial personnel C in enterprise collaboration.

[0056] 5. Data privacy and security protection

[0057] In best practices, privacy is protected throughout the sharing process:

[0058] • The prompt word description information D can mark sensitive data and transmit it through automatic encryption;

[0059] • Personalized processing supports automatic masking or replacement of sensitive information;

[0060] •The sharing link uses HTTPS or other encryption protocols to ensure data security.

[0061] 6. Modular implementation of the device

[0062] The prompt word sharing device adopts a modular design to facilitate function expansion:

[0063] •Storage module: supports classified storage of multiple prompt word sets, such as by scenario (e-commerce, law, education) or content type (text, picture).

[0064] •Forwarding module: supports efficient transmission and ensures that the large model M can receive the prompt word set instantly.

[0065] •Processing module: Dynamically modify the prompt word set based on preset rules or real-time input. Modes for Carrying Out the Invention

[0066] 1. E-commerce scenario: product returns

[0067] 1.1 Generation of prompt word set P

[0068] •Sharer A generates a set of prompt words P, which includes prompt word 1: "Return the mobile phone purchased yesterday from the e-commerce platform."

[0069] • Prompt Description Information D: This control instruction requires the large model to automatically activate the camera function of recipient B's phone during operation to take a photo of the returned product. This also includes the required resolution and content of the photo, such as "front view of the product" or "photo of the packaging integrity."

[0070] 1.2 Sharing Methods

[0071] •Sharer A generates sharing content through a link and sends it to recipient B through social software, without the need to explicitly install or download any software.

[0072] 1.3 Receiver B processes the prompt word set P

[0073] •After the receiver B clicks the link, the prompt word set P and description information D are automatically loaded into the large model M platform.

[0074] •During operation, the large model calls the camera function of the receiver B’s mobile phone through the description information D, automatically waking up the camera application without manual operation.

[0075] •Receiver B follows the model’s prompts to complete the product photo shoot, and the photos are automatically uploaded to model M.

[0076] 1.4 Large Model M Processes Prompt Word Set P

[0077] •Model M retrieves the target product order information through the order interface of the e-commerce platform.

[0078] •Use the photos uploaded by recipient B in combination with the order data to generate a return application document that meets the return requirements.

[0079] •If additional materials are required (such as photos of invoices), the model will continue to call up the camera function and prompt the user to take photos of the relevant documents.

[0080] 1.5 Return result generation

[0081] •The large model M returns the return application results, including the return logistics order number and logistics requirements.

[0082] •If additional material is needed, the model generates new prompt words and continues to interact with the user device.

[0083] 1.6 Technical Highlights and Advantages

[0084] • Intelligent interaction: Directly call the hardware of the user device through description information D to achieve seamless integration of prompt words and operating environment.

[0085] •Seamless process: Receiver B only needs to complete the shooting action, and the rest of the steps are completed automatically by the large model.

[0086] •Privacy and security: Descriptive information D ensures the security of device interactions through permission management, such as only calling the camera function and not storing user photos.

[0087] 1.7 Extended Application

[0088] • Personalization before sharing: Sharer A customizes the prompt words according to the specific product return requirements, such as adding "photo of the back of the product and serial number."

[0089] • Personalization after sharing: Recipient B can add additional explanation, such as "Reason for return: The product has obvious scratches."

[0090] •Specified use of results: The return application results can be sent to sharer A simultaneously to confirm the return progress.

[0091] 2. Enterprise scenario: Law firm billing collaboration

[0092] 2.1 Generation of prompt word set P

[0093] •Lawyer A generates prompt word 1: "The client has a contract dispute involving 500,000 yuan, and the estimated resolution time is 3 months."

[0094] • Prompt 2: "Depending on the complexity of the case, the suggested fee range is 50,000 to 80,000 yuan."

[0095] •Descriptive information D includes the basis for fee recommendations (such as case type, time, and complexity) and sensitive information shielding rules (such as the client's financial status is for the lawyer's reference only).

[0096] 2.2 Sharing Methods

[0097] •Lawyer A shares the prompt word set P and description information D to client communicator B in the form of push through the law firm's internal collaboration platform.

[0098] 2.3 Customer Communicator B processes prompt word set P

[0099] •B loads the prompt word set P into the large model M and generates a specific cost communication plan.

[0100] •If the customer is an overseas customer, B can add a new prompt: "The customer wishes to pay via a cross-border payment method. Please generate the payment process and related precautions."

[0101] 2.4 Large Model M Processes Prompt Word Set P

[0102] •Model M generates a detailed cost breakdown, including terms of service, payment methods, and notes.

[0103] •For newly added prompt words, the model generates corresponding tax process suggestions and simultaneously expands the prompt words to improve the payment plan.

[0104] 2.5 Specified use of results

[0105] •The expense details generated by model M are sent directly to customer communicator B and synchronized to financial officer C for review and tracking of payment status.

[0106] 2.6 Extended Application

[0107] • Personalization before sharing: Lawyer A can block sensitive information such as client property details.

[0108] • Personalization after sharing: B can add prompt words, such as "Generate installment payment options and precautions."

[0109] • Collaborative logging: The system generates a log of prompt word modifications and result usage for easy auditing and management.

[0110] 3. Medical scenarios

[0111] •Prompt word set P: "Upload physical examination report and generate health analysis report".

[0112] •Descriptive information D includes patient age, gender, and other information.

[0113] •Sharing person A (such as a doctor) shares a set of prompt words P with patient B. After patient B uploads a picture, the big model generates a health report.

[0114] 4. Educational scenarios

[0115] •Cue word set P: “Develop personalized learning plans for middle school students.”

[0116] •Descriptive information D includes learning topics and knowledge point coverage.

[0117] •Sharer A (such as a teacher) shares the prompt words through a link, and student B modifies the content according to needs and generates a learning plan.

[0118] 5. User-generated content (UGC) platform scenarios

[0119] •Prompt word set P: “Generate a short video script on the theme of future technology”.

[0120] •Description information D includes video style, length, music, etc.

[0121] •Sharer A shares the prompt words to the large model M to generate a creative script.

[0122] 6. Game development scenarios

[0123] •Prompt word set P: "Design a medieval-style game level, including terrain and NPC settings."

[0124] •Sharer A shares the prompt words to receiver B, who adjusts the NPC settings according to needs and completes the level design. Industrial Applicability

[0125] The present invention provides a method and device for sharing prompt words, which has broad industrial applicability, specifically in the following aspects:

[0126] 1. Cross-industry versatility

[0127] •The present invention is applicable to multiple industry scenarios, including but not limited to e-commerce, law, medical care, education, user-generated content (UGC) platforms, and game development.

[0128] •In various industries, the present invention improves the efficiency and accuracy of data interaction through prompt word sharing and processing, and significantly reduces the cost of manual participation.

[0129] 2. Intelligent operation

[0130] •By combining the prompt word description information D with the artificial intelligence large model, the present invention can automatically arouse the user device function (such as camera) and realize the full process automation of the task.

[0131] •For example, in an e-commerce scenario, the present invention enables users to quickly complete the product return process without having to manually search for order information or fill out complex forms, greatly improving the user experience.

[0132] 3. Improve collaboration efficiency

[0133] • In enterprise collaboration, this invention enables efficient information flow between different positions through seamless sharing and personalized adjustment of prompt words. For example, in a law firm billing scenario, collaboration between lawyers, client liaisons, and financial personnel becomes more precise and efficient, significantly improving overall business processing efficiency.

[0134] 4. Low-threshold applications

[0135] • There are various ways to share, including QR codes, links, push notifications, etc. Users can receive and use prompt words without explicitly installing or downloading any software.

[0136] •The prompt word sharing process seamlessly connects the user's operating environment and the artificial intelligence large model, reducing technical complexity and facilitating wide deployment in various hardware and software environments.

[0137] 5. Strong scalability

[0138] •The present invention supports personalized modification and dynamic expansion of prompt words, such as adding new prompt words or adjusting the content of the original prompt words according to actual needs after sharing.

[0139] •In different scenarios, the results generated by the large model can be flexibly distributed to sharers, recipients or third parties, helping to meet diverse business needs.

[0140] 6. Reduce operating costs

[0141] • This invention leverages the intelligent capabilities of large models to reduce manual intervention and significantly lower enterprise operating costs. For example, in medical scenarios, doctors can quickly generate health reports using prompts, reducing the workload of manual data processing and analysis.

[0142] 7. Data security and privacy protection

[0143] •Sensitive data contained in the prompt word description information D can be protected by shielding or encryption technology before sharing to avoid the risk of data leakage.

[0144] •During the entire process of prompt word sharing and execution, the present invention ensures the security of user device interaction through permission management and encryption protocols, which meets the strict requirements of modern industry for data privacy and security.

[0145] 8. Huge market potential

[0146] •With the widespread application of large artificial intelligence models, this invention can help more users fully tap the potential of large models, improve user experience, and has great market promotion value.

[0147] •This invention can also promote the standardization and modularization of artificial intelligence services, achieve seamless collaboration between different platforms and users, and further promote the development of the artificial intelligence industry.

[0148] In summary, the present invention meets the needs for efficiency, intelligence, and security in multiple industry scenarios through the deep integration of prompt word sharing and large models, and has broad industrial applicability and market value.

Claims

1. A method for sharing prompt words, characterized in that: Broadly speaking, prompt words consist of purpose and reference information content, including but not limited to large-scale model input information such as text, pictures, voice, and video. The sharing steps include: S1 "Sharer A" shares "generalized prompt word set P" with "receiving sharer B", "generalized prompt word set P" consists of one or more prompt words; S2 "generalized prompt word set P" will be received by the designated "artificial intelligence large model M"; The "artificial intelligence large model M" specified by S3 processes the prompt word P and gives the result of the model.

2. A method for sharing prompt words according to claim 1, characterized in that: S4 sharing methods include but are not limited to QR codes, links, and push notifications. The sharing process does not require manual copying of all or part of the prompt words, and there is no explicit download or installation process.

3. A method for sharing prompt words according to claim 1, characterized in that: S5 The information shared by "sharer A" to "receiver B" will also include "prompt word description information D". The content of "prompt word description information D" includes but is not limited to the specified model, author, price, evaluation, judgment conditions and control instructions related to the prompt word.

4. A method for sharing prompt words according to claim 1, characterized in that: S6 personalizes the "prompt word set P" into pre-sharing personalization and post-sharing personalization.

5. A method for sharing prompt words according to claim 1, characterized in that: S7: Personalized modification of "prompt word set P", including modifying a prompt word and deleting a prompt word.

6. A method for sharing prompt words according to claim 1, characterized in that: S8 Personalized modification of "prompt word set P", including adding a prompt word.

7. A method for sharing prompt words according to claim 1, characterized in that: S9 can specify the user of the results given by the "artificial intelligence big model M", and the user includes but is not limited to "sharer A", "sharer B" or a third party.

8. A device for sharing prompt words, characterized in that: Includes the following modules: D1 storage module is responsible for storing the "prompt word set PA" of "sharer A", the "prompt word set PB" received by "sharer B", the address of the designated "artificial intelligence big model M", the results returned after processing by the big model, and the intermediate address generated during sharing; The D2 forwarding module forwards the PA in D1 to the PB in D1 for storage, forwards the PB to the M in D1 to request large model processing, and retrieves the results processed by the M in D1 for storage.

9. The device for sharing prompt words according to claim 8, characterized in that: D1 will also include the "prompt word description information D" corresponding to PA and PB.

10. The device for sharing prompt words according to claim 8, characterized in that: Includes the following modules: The D3 processing module modifies, deletes and adds PA and PB in D1 as required.

Images