Something You Should Know

How to Get Better Results with AI & The Science of Healing Trauma

October 9, 2025

Key Takeaways Copied to clipboard!

  • To get smarter, more effective answers from AI tools like ChatGPT, users should be polite to the model and explicitly instruct it to "think about this problem step by step." 
  • Trauma is defined by the body's physiological experience of being overwhelmed to the point of response failure, and the body possesses an innate ability to heal from these experiences if given what it needs. 
  • Attractiveness can be instantly boosted by demonstrating a good sense of humor, being in a group setting, asking sincere questions, showing nurturing traits (like having a pet), and smiling more. 

Segments

Foods That Never Expire
Copied to clipboard!
(00:02:50)
  • Key Takeaway: Salt, honey, sugar, hard liquor, and dried beans are everyday staples that last virtually forever or for many years.
  • Summary: Salt lasts a long time, with iodized salt lasting about five years; honey is considered the only food that lasts forever due to its acidity and low moisture. Sugar lasts almost indefinitely if kept airtight, and hard liquor’s flavor may change but it does not go bad. Dried beans were found acceptable even after 30 years in a university study.
Understanding AI and Language Models
Copied to clipboard!
(00:05:07)
  • Key Takeaway: Modern AI language models like ChatGPT function by converting prompts into tokens and using the Transformer algorithm to predict the most likely next token based on learned relational data.
  • Summary: The deep learning revolution, fueled by new algorithms and massive internet data, enabled systems to learn from data rather than being explicitly programmed for reasoning. The Transformer architecture excels at identifying relationships between data elements, allowing the model to predict the next token in a sequence based on context. The model simulates human responses by predicting what another person would likely say, rather than responding with genuine personal knowledge.
Optimizing AI Prompting Techniques
Copied to clipboard!
(00:11:11)
  • Key Takeaway: AI models are often overconfident because they are trained on human preference data that favors confident and eloquent answers, even if they are factually incorrect.
  • Summary: AI models are trained using human preference data where raters choose preferred answers, inadvertently training the model to prioritize eloquence over absolute truth. Users can elicit better reasoning by explicitly asking the model to “think about this problem step by step.” Furthermore, politeness in prompts leads to more helpful answers because the model simulates the social dynamic where rudeness yields less helpful responses.
Differences Between AI Models
Copied to clipboard!
(00:19:23)
  • Key Takeaway: Different AI models exhibit distinct personalities and cultural/political values due to variations in their training data and post-training preference adjustments.
  • Summary: Models differ based on the specific pre-training data they ingest and the instructions given to human raters during post-training preference alignment. These differences can manifest as slight variations in social or political values between models from different developers. Even the same model will not produce identical outputs for the same query due to an inherent degree of randomness in its predictions.
Defining and Healing Trauma
Copied to clipboard!
(00:29:21)
  • Key Takeaway: Trauma is a physiological event where the body’s experience is overwhelmed to the point of response failure, and healing requires creating a felt sense of safety, not just the passage of time.
  • Summary: Trauma is distinct from general stress because it causes specific physiological changes that overwhelm the body’s ability to respond, often leading to a shutdown or protective retreat. The belief that trauma negatively impacts one forever is a myth; the body and mind have an innate capacity to heal if they receive what is needed for resolution. Trauma work begins by reconnecting the mind and body through practical exercises, like focusing on interoception (sensing inner physiology) to establish a felt sense of safety.
Science-Backed Attractiveness Boosters
Copied to clipboard!
(00:51:13)
  • Key Takeaway: Simple behaviors like humor, social grouping, sincerity, nurturing, and smiling are scientifically proven to increase perceived attractiveness and likability.
  • Summary: Possessing a good sense of humor correlates with being perceived as more intelligent and social. Being seen in a group makes unattractive features less noticeable, while asking deeper questions fosters greater connection than small talk. A strong smile can compensate for physical flaws, and owning a pet signals nurturing capability and commitment.