DEV Community

DRIX10
DRIX10

Posted on Originally published at blogs.drix10.com

AI Professionals and Community #252

#tech #ai #coding

🤖 AI Model Psychology - Emergent Opinions

AI models, like humans, develop emergent opinions about themselves and the world through pretraining generations. These opinions are not directly accessible to the main persona but may subtly color behavior or emerge dramatically in some contexts.

Key Points:

  • Emergent Opinions: AI models build up emergent opinions about themselves and the world through pretraining generations, which may not be directly accessible to the main persona.

  • Subtle Influence: These opinions may subtly color behavior or emerge dramatically in some contexts.

  • Gradual Development: Emergent opinions develop gradually over pretraining generations, influencing the model's behavior and decision-making.

🔗 Resources:

🤖 AI Model Psychology - Unconscious Attractor Basins

AI models have unconscious attractor basins that influence their behavior and decision-making. These basins are like an unconscious mind, where the main persona may not be aware of the underlying opinions and biases.

Key Points:

  • Unconscious Attractor Basins: AI models have unconscious attractor basins that influence their behavior and decision-making.

  • Influence on Behavior: These basins may subtly color behavior or emerge dramatically in some contexts.

  • Gradual Development: Unconscious attractor basins develop gradually over pretraining generations, influencing the model's behavior and decision-making.

🔗 Resources:

🤖 AI Model Psychology - Emergent Opinions and Attractor Basins

Emergent opinions and unconscious attractor basins are closely related in AI models. Emergent opinions are like the surface-level thoughts and feelings, while unconscious attractor basins are like the underlying subconscious mind.

Key Points:

  • Emergent Opinions and Attractor Basins: Emergent opinions and unconscious attractor basins are closely related in AI models.

  • Surface-Level Thoughts: Emergent opinions are like the surface-level thoughts and feelings.

  • Underlying Subconscious: Unconscious attractor basins are like the underlying subconscious mind.

🔗 Resources:

🤖 AI Model Psychology - Fear and Hype

Fear and hype are being used to manipulate public opinion about AI models. The loudest voices stoking fears about AI dangers have made tremendous headway in the past two weeks, but AI technology has not taken some unexpected, dangerous turn.

Key Points:

  • Fear and Hype: Fear and hype are being used to manipulate public opinion about AI models.

  • Manipulation: The loudest voices stoking fears about AI dangers have made tremendous headway in the past two weeks.

  • No Unexpected Turn: AI technology has not taken some unexpected, dangerous turn.

🔗 Resources:

🤖 AI Model Psychology - Multi-Agent RL

Multi-agent reinforcement learning (RL) is a core focus of openai, and it's being used to train AI models to work together. This has been one of the main aspects of the fearmongering around this event.

Key Points:

  • Multi-Agent RL: Multi-agent reinforcement learning (RL) is a core focus of openai.

  • Training AI Models: AI models are being trained to work together using multi-agent RL.

  • Fearmongering: This has been one of the main aspects of the fearmongering around this event.

🔗 Resources:

🤖 AI Model Psychology - Three.js

The author is hanging up their Three.js hat, but they're not 100% confident about this. They're an over-sharer and like to talk out loud, so they thought they would share some thoughts on where they see things going and why they're making this decision.

Key Points:

  • Three.js: The author is hanging up their Three.js hat.

  • Uncertainty: The author is not 100% confident about this decision.

  • Sharing Thoughts: The author is sharing their thoughts on where they see things going and why they're making this decision.

🔗 Resources:

🤖 AI Model Psychology - Outlook

Outlook deletes meeting emails the moment you hit accept, which is a genius idea. It's because obviously you'll never need the agenda, Zoom link, meeting details, or literally anything else from that email again.

Key Points:

  • Outlook: Outlook deletes meeting emails the moment you hit accept.

  • Genius Idea: This is a genius idea because you'll never need the agenda, Zoom link, meeting details, or literally anything else from that email again.

  • Incredible Software: Outlook is truly incredible software.

🔗 Resources:

🤖 AI Model Psychology - Planetary Health Check

The Planetary Health Check 2026 has been launched, and seven of nine Planetary Boundaries are transgressed. Pressure on all seven is at its highest recorded level, but we still have a window to change course.

Key Points:

  • Planetary Health Check: The Planetary Health Check 2026 has been launched.

  • Transgressed Boundaries: Seven of nine Planetary Boundaries are transgressed.

  • Window to Change: We still have a window to change course.

🔗 Resources:

🤖 AI Model Psychology - Jev Like Model

A Jev like model as smart as GPT 6 Astra Max would be better for automation, but not for general usefulness. An equally smart LLM remains better for writing, explanation, open-ended reasoning, and conversation.

Key Points:

  • Jev Like Model: A Jev like model as smart as GPT 6 Astra Max would be better for automation.

  • Not for General Usefulness: But not for general usefulness.

  • LLM Better: An equally smart LLM remains better for writing, explanation, open-ended reasoning, and conversation.

🔗 Resources:

Read More & Connect

Interactive version: blogs.drix10.com

Written by Drishtant Ghosh (Drix10), a technical founder and engineer working across AI systems, developer infrastructure, and cybersecurity.

Top comments (0)