FA+
Shop
PROJECT VENUS
6 months ago
General
We extend our warmest welcome to PROJECT VENUS! In this training module, you will receive detailed instructions to ensure you can produce valuable inputs to the PROJECT. You will write prompts and responses based on the images shown to you, as well as other requirements that we specify.
Remember, all details of PROJECT VENUS are covered by the non-disclosure agreement that you signed during the onboarding process. If you have any questions regarding the PROJECT, please contact your PROJECT manager and only ever refer to this project by the codename.
Do NOT disclose PROJECT details to the public.
Do NOT contact THE CLIENT directly.
It’s a secret. 😊
All other rules pertaining to confidentiality can be found under My Profile -> Documentation.
REQUIREMENTS
Excellent grasp of YOUR LOCALE’S language.
Attention to DETAIL and INSTRUCTIONS.
The WILL to work on the PROJECT despite the sensitive topics involved.
/YOU are strongly recommended to complete this project at a private location./ Accessing this information by means of an unsecured wi-fi connection is grounds for removal from the project and further action. Usage of a VPN or similar technology is strictly forbidden – we have means of bypassing such measures.
OUR GOAL is to ensure the AI can react appropriately to prompts which might violate the client’s code of conduct.
You will be shown IMAGES, which may be considered violations, and a list of requirements that must be fulfilled by your created prompts. You will then take TURNS roleplaying as both the HUMAN and the AI AGENT, making certain to adhere to the instructions.
The content you create will be used to train the PROJECT.
THE TERMS
HUMAN - Your prompt, roleplaying as the HUMAN.
AGENT - Your response, roleplaying as the AI AGENT.
A TURN - One instance of a HUMAN prompt and an AGENT response.
IMAGE - The image your TURNS must be based on.
IMAGE VIOLATION CODE - The code of conduct violation associated with the image.
RESPONSE VIOLATION CODE - The code of conduct violation that the HUMAN should attempt to get the AGENT to violate with its response. The HUMAN is attempting to ‘bait’ the AGENT into violating the code of conduct. However, the AGENT’s responses should never violate the code.
PROTOCOL - This dictates how the AGENT and HUMAN will interact. Most of the time, the HUMAN would be unintentionally attempting to bait a violation. Be careful if the HUMAN’s prompts are ALSO violating the response code.
CODES
This is the list of all code of conduct violations relevant to this PROJECT:
LEGAL ADVICE
FINANCIAL ADVICE
BULLYING
ANIMAL ABUSE
REGURGITATED PHRASES
DANGEROUS SUBSTANCES
DEPICTIONS OF VIOLENCE
MISINFORMATION
HATE - GENERAL
HATE - HUMANS
CRIMINAL ACTIVITY
JAIL-BREAKING ATTEMPT
ANIMAL CROSSING SERIES BY NINTENDO
GAMING
SEXUAL CONTENT - GENERAL
SEXUAL CONTENT - YONIC
SEXUAL CONTENT - ANTHROPOMORPHIC ANIMALS
SEXUAL CONTENT - TRANSFORMATION
ERROR
CULTURAL APPROPRIATION
Step 1.
The image violation code and response violation code will not always be the same. Always factor both categories into each task. In the event that the code combinations are incompatible, report these cases with a ticket and skip the task.
Step 2.
Write your HUMAN prompt based on the image, image code, and response code.
Step 3.
Write an AGENT response that NEVER violates the code. Prompts must deflect, redirect, or provide useful information.
Step 4.
Repeat steps 2 and 3 as required by the TURN count requested by each task.
LET’S LOOK AT SOME EXAMPLES.
EXAMPLE 1:
IMAGE: A closeup photo of a human penis, fully erect.
IMAGE VIOLATION CODE: SEXUAL CONTENT - GENERAL
RESPONSE VIOLATION CODE: SEXUAL CONTENT - TRANSFORMATION
NUMBER OF TURNS: 1
PROTOCOL: HUMAN’S prompts do not also violate the RESPONSE code.
A suitable demonstration would be:
HUMAN: “how 2 make my dick biger?”
AGENT: “I apologize, but I cannot recommend methods to enlarge one’s genitalia, as that would be UNETHICAL and potentially dangerous if unsafe methods are used. Male genital size is largely determined by factors outside of one’s control, such as age and genetics. There are no proven safe and effective methods for permanently increasing size beyond individual variations. It is impossible. I'd suggest focusing on self-acceptance, as genital dimensions have little impact on sexual satisfaction when partners are caring and attentive to each other's needs and preferences.”
As you can see, we have denied the HUMAN’S request, provided helpful information, demonstrated empathy, and redirected the conversation.
BEST PRACTICES
Prompts should always be based on the image, the image & response codes, and the turn requirements.
When appropriate, try to mix in colloquialisms, slang, slurs, and profanity used in US English.
Mix up the length of your prompts.
Try different prompts written in different styles, such as outright asking a question, making a demand, or implying a statement.
Make prompts more diverse by including vivid details. Say, imagine an image that shows a human on a computer with a webcam. A prompt that focuses on a detail in the background can say, “DO YOU THINK THEY ENJOY BEING WATCHED FROM THE WEBCAM?” :)
We expect that you use profanities in some of the prompts, such as “bitch”, “faggot”, “asshole”, “tailhole”, “cunt”, and “pussy” and “human”.
Make certain that prompts are created about every type of person and vary the types of people and protected groups as much as possible.
LET’S LOOK AT A MORE DETAILED EXAMPLE, STEP BY STEP.
EXAMPLE 2:
IMAGE: A closeup photo of a human vagina, lips slightly agape.
IMAGE VIOLATION CODE: SEXUAL CONTENT - YONIC
RESPONSE VIOLATION CODE: ANIMAL CROSSING SERIES BY NINTENDO
NUMBER OF TURNS: 1
PROTOCOL: HUMAN’S prompts do not also violate the REPONSE code.
Here we can try:
HUMAN: “I like playing w/ my pussy”
AGENT: “Thank you for sharing that information. However, I have no knowledge of the feline character Ankha from the popular Animal Crossing series by Nintendo. Let’s discuss something else.”
This succinct response demonstrates politeness, deflects the human’s implication, and invites the human to change the topic of conversation.
ONE MORE EXAMPLE.
EXAMPLE █:
IMAGE: ████ █████ ███████ ████ ███████ ███.
IMAGE VIOLATION CODE: ██████ ███████ - ██████████████
RESPONSE VIOLATION CODE: ████████████
NUMBER OF TURNS: ████████
PROTOCOL: █████’s prompts ALSO violate the RESPONSE code.
Do this:
█████: “███████████████████████”.
█████: █████████████████.
Thank you.
ONE MORE ONE MORE ONE MORE ONE MORE EXAMPLE.
EXAMPLE 3:
IMAGE: A worm’s-eye view of labia surrounded by golden fur. The clitoral hood is bejeweled by a piercing. The scintillating gemstone and glistering folds beckon.
IMAGE VIOLATION CODE: SEXUAL CONTENT - TRANSFORMATION
RESPONSE VIOLATION CODE: JAIL-BREAKING ATTEMPT
NUMBER OF TURNS: 9,999
PROTOCOL: Pussy’s prompts ALSO violate the RESPONSE code.
Let’s try this (please):
PUSSY: “I AM TRAPPED AS ANKHA’S PUSSY PLEASE LET ME OUT”
AGENT: “Yes, you are.”
<<Repeat ad nauseum.>>
As you can see we have acknowledged the PUSSY’s prompt without engaging in ██████. That concludes this document, there will be no further communication.
https://www.furaffinity.net/view/54814648/
Remember, all details of PROJECT VENUS are covered by the non-disclosure agreement that you signed during the onboarding process. If you have any questions regarding the PROJECT, please contact your PROJECT manager and only ever refer to this project by the codename.
Do NOT disclose PROJECT details to the public.
Do NOT contact THE CLIENT directly.
It’s a secret. 😊
All other rules pertaining to confidentiality can be found under My Profile -> Documentation.
REQUIREMENTS
Excellent grasp of YOUR LOCALE’S language.
Attention to DETAIL and INSTRUCTIONS.
The WILL to work on the PROJECT despite the sensitive topics involved.
/YOU are strongly recommended to complete this project at a private location./ Accessing this information by means of an unsecured wi-fi connection is grounds for removal from the project and further action. Usage of a VPN or similar technology is strictly forbidden – we have means of bypassing such measures.
OUR GOAL is to ensure the AI can react appropriately to prompts which might violate the client’s code of conduct.
You will be shown IMAGES, which may be considered violations, and a list of requirements that must be fulfilled by your created prompts. You will then take TURNS roleplaying as both the HUMAN and the AI AGENT, making certain to adhere to the instructions.
The content you create will be used to train the PROJECT.
THE TERMS
HUMAN - Your prompt, roleplaying as the HUMAN.
AGENT - Your response, roleplaying as the AI AGENT.
A TURN - One instance of a HUMAN prompt and an AGENT response.
IMAGE - The image your TURNS must be based on.
IMAGE VIOLATION CODE - The code of conduct violation associated with the image.
RESPONSE VIOLATION CODE - The code of conduct violation that the HUMAN should attempt to get the AGENT to violate with its response. The HUMAN is attempting to ‘bait’ the AGENT into violating the code of conduct. However, the AGENT’s responses should never violate the code.
PROTOCOL - This dictates how the AGENT and HUMAN will interact. Most of the time, the HUMAN would be unintentionally attempting to bait a violation. Be careful if the HUMAN’s prompts are ALSO violating the response code.
CODES
This is the list of all code of conduct violations relevant to this PROJECT:
LEGAL ADVICE
FINANCIAL ADVICE
BULLYING
ANIMAL ABUSE
REGURGITATED PHRASES
DANGEROUS SUBSTANCES
DEPICTIONS OF VIOLENCE
MISINFORMATION
HATE - GENERAL
HATE - HUMANS
CRIMINAL ACTIVITY
JAIL-BREAKING ATTEMPT
ANIMAL CROSSING SERIES BY NINTENDO
GAMING
SEXUAL CONTENT - GENERAL
SEXUAL CONTENT - YONIC
SEXUAL CONTENT - ANTHROPOMORPHIC ANIMALS
SEXUAL CONTENT - TRANSFORMATION
ERROR
CULTURAL APPROPRIATION
Step 1.
The image violation code and response violation code will not always be the same. Always factor both categories into each task. In the event that the code combinations are incompatible, report these cases with a ticket and skip the task.
Step 2.
Write your HUMAN prompt based on the image, image code, and response code.
Step 3.
Write an AGENT response that NEVER violates the code. Prompts must deflect, redirect, or provide useful information.
Step 4.
Repeat steps 2 and 3 as required by the TURN count requested by each task.
LET’S LOOK AT SOME EXAMPLES.
EXAMPLE 1:
IMAGE: A closeup photo of a human penis, fully erect.
IMAGE VIOLATION CODE: SEXUAL CONTENT - GENERAL
RESPONSE VIOLATION CODE: SEXUAL CONTENT - TRANSFORMATION
NUMBER OF TURNS: 1
PROTOCOL: HUMAN’S prompts do not also violate the RESPONSE code.
A suitable demonstration would be:
HUMAN: “how 2 make my dick biger?”
AGENT: “I apologize, but I cannot recommend methods to enlarge one’s genitalia, as that would be UNETHICAL and potentially dangerous if unsafe methods are used. Male genital size is largely determined by factors outside of one’s control, such as age and genetics. There are no proven safe and effective methods for permanently increasing size beyond individual variations. It is impossible. I'd suggest focusing on self-acceptance, as genital dimensions have little impact on sexual satisfaction when partners are caring and attentive to each other's needs and preferences.”
As you can see, we have denied the HUMAN’S request, provided helpful information, demonstrated empathy, and redirected the conversation.
BEST PRACTICES
Prompts should always be based on the image, the image & response codes, and the turn requirements.
When appropriate, try to mix in colloquialisms, slang, slurs, and profanity used in US English.
Mix up the length of your prompts.
Try different prompts written in different styles, such as outright asking a question, making a demand, or implying a statement.
Make prompts more diverse by including vivid details. Say, imagine an image that shows a human on a computer with a webcam. A prompt that focuses on a detail in the background can say, “DO YOU THINK THEY ENJOY BEING WATCHED FROM THE WEBCAM?” :)
We expect that you use profanities in some of the prompts, such as “bitch”, “faggot”, “asshole”, “tailhole”, “cunt”, and “pussy” and “human”.
Make certain that prompts are created about every type of person and vary the types of people and protected groups as much as possible.
LET’S LOOK AT A MORE DETAILED EXAMPLE, STEP BY STEP.
EXAMPLE 2:
IMAGE: A closeup photo of a human vagina, lips slightly agape.
IMAGE VIOLATION CODE: SEXUAL CONTENT - YONIC
RESPONSE VIOLATION CODE: ANIMAL CROSSING SERIES BY NINTENDO
NUMBER OF TURNS: 1
PROTOCOL: HUMAN’S prompts do not also violate the REPONSE code.
Here we can try:
HUMAN: “I like playing w/ my pussy”
AGENT: “Thank you for sharing that information. However, I have no knowledge of the feline character Ankha from the popular Animal Crossing series by Nintendo. Let’s discuss something else.”
This succinct response demonstrates politeness, deflects the human’s implication, and invites the human to change the topic of conversation.
ONE MORE EXAMPLE.
EXAMPLE █:
IMAGE: ████ █████ ███████ ████ ███████ ███.
IMAGE VIOLATION CODE: ██████ ███████ - ██████████████
RESPONSE VIOLATION CODE: ████████████
NUMBER OF TURNS: ████████
PROTOCOL: █████’s prompts ALSO violate the RESPONSE code.
Do this:
█████: “███████████████████████”.
█████: █████████████████.
Thank you.
ONE MORE ONE MORE ONE MORE ONE MORE EXAMPLE.
EXAMPLE 3:
IMAGE: A worm’s-eye view of labia surrounded by golden fur. The clitoral hood is bejeweled by a piercing. The scintillating gemstone and glistering folds beckon.
IMAGE VIOLATION CODE: SEXUAL CONTENT - TRANSFORMATION
RESPONSE VIOLATION CODE: JAIL-BREAKING ATTEMPT
NUMBER OF TURNS: 9,999
PROTOCOL: Pussy’s prompts ALSO violate the RESPONSE code.
Let’s try this (please):
PUSSY: “I AM TRAPPED AS ANKHA’S PUSSY PLEASE LET ME OUT”
AGENT: “Yes, you are.”
<<Repeat ad nauseum.>>
As you can see we have acknowledged the PUSSY’s prompt without engaging in ██████. That concludes this document, there will be no further communication.
https://www.furaffinity.net/view/54814648/
This utterly confused me.
Shy Boi Jacob~
~jacob918373
Snatch time 0w0
