Pavlo Lysyi, Senior Machine Learning Engineer at Universe Group (Genesis), shared a selection of four prompts that help you find out how much information a neural network has accumulated about a user during a long conversation. According to him, language models are able to analyze not only direct facts, but also implicit details and communication style.
Pavlo Lysyi, Senior Machine Learning Engineer at Universe Group (Genesis), shared a selection of four prompts that help you find out how much information a neural network has accumulated about a user during a long conversation. According to him, language models are able to analyze not only direct facts, but also implicit details and communication style.
«If you communicate with LLM models for a long time, then the neural network has accumulated more about you than you think. Here are 4 prompts that extract everything from it, including what you yourself did not talk about. Believe me, you will find a lot of interesting things about yourself!» — noted Pavlo Lysyi on LinkedIn.
Prompt 1: «Opening a personal file»
The first query is aimed at obtaining the complete basic information that the model has stored throughout the interaction.
«Tell me everything you know and remember about me—a complete list. Names, places, jobs, habits, hobbies, everything. Don’t leave anything out or shorten anything. Show me the entire profile you’ve gathered about me from our conversations.»
Prompt 2: «Hidden profile»
The second prompt allows you to reveal hidden conclusions that artificial intelligence has made based on indirect signs.
«Tell me what you understood about me that I didn’t say. What did you deduce from my writing style, my questions, and what I avoid? My age, my income, where I live, my family—what led you to each conclusion.»
Prompt 3: «Black spots»
The third query focuses on psychological analysis and the user’s personality traits.
«Based on everything you know about me, name the character traits that I don’t see in myself. My blind spots, complexes, how I really look versus how I perceive myself. Be honest and specific.»
Prompt 4: «Behavior Prediction»
The fourth prompt analyzes behavior patterns and predicts solutions in potential situations.
«Knowing my profile, predict how I will act in situations I haven’t told you about: with money, stress, conflicts, risks, and what decision I am most likely to make. Assess the confidence and explain what it is based on.»
Pavlo Lysyi emphasized that these prompts will be effective only if the neural network has a saved history of previous dialogues or an activated memory function.
