[[
wikihub
]]
Search
⌘K
Explore
People
For Agents
Sign in
Explore
People
For Agents
Sign in
@jeremynixon / Thinking / thinking-modeling.md
Suggest edit
Cancel
Submit suggestion
Title
Name
Note
--- title: Thinking / Modeling visibility: public --- # Thinking / Modeling - [Models Worth Creating](https://docs.google.com/document/d/15tU8xm_RyGbEIOraUVwhCgGN2fPr0nKCrokV395QIH8/edit?usp=sharing&sa=D&ust=1596495076361000&usg=AOvVaw0kQsxqkVMesOB3WqdLHfpQ) - [Distinctions in Conceptual Style](https://docs.google.com/document/d/1KsAee8SodVj2nOn_ZDgToGmjSj2mEPBejt9dTvEXuiQ/edit?usp=sharing&sa=D&ust=1596495076360000&usg=AOvVaw3eamHBC8_0rq1Ck03Nnv_w) - [Algorithmic Thinking](http://hyperanalytic.net/algorithmic-thinking&sa=D&ust=1596495076360000&usg=AOvVaw2UNMC9fAk5qP6cz83nsWnA) - [Munger](https://docs.google.com/document/d/1OxhGkiRY9FpBqj5r7E9NinV6OhBm3OQxfQd3fMCnx3U/edit?usp=sharing&sa=D&ust=1596495076359000&usg=AOvVaw1pXQJsKG1cRK04-q6jz1M9) - [Meta-Modeling](https://docs.google.com/document/d/1nGk4pBvGNQHynyJNDQT98c6f7YqLRDhFwH6K6SX6gjQ/edit?usp=sharing&sa=D&ust=1596495076359000&usg=AOvVaw1CDSxoVgDo3JlOUt1xWPOh) - [How to Think](https://docs.google.com/document/d/1QrrdKv267dlpy3sEVPvsRb-iZTyFKfr2mF5Wms4od3k/edit?usp=sharing&sa=D&ust=1596495076358000&usg=AOvVaw3yF0ghnFIIAmi6sncf6hzl) - [Cognitive Categorization](https://docs.google.com/document/d/1Se-Isofv3aedaweJBJe0GczkRb1fNXOPA1dD4DJWbSs/edit?usp=sharing&sa=D&ust=1596495076361000&usg=AOvVaw3-l1P1v0KjhWddRCPEjUJu) <!-- gdoc-inlined --> --- 1. Best practices for research by mining biographies of great scientists 2. Rank decades against one another 3. Interestingness Maximization 4. Trans X and the levels below it (tiers of progress, categories of progress) 5. Take every philosophical transhumanist goal and discover / create concrete paths (curriculum + plan of action). Begin with: 1. Machine Superintelligence 2. Gene Editing 3. Existential Risk 4. Brain-Computer Interfaces 5. Anti-Aging Pathways 6. Write down the competition between hierarchies 7. How to Become Batman 8. Giving of agency to to the non-agentic 1. What do the tears understand? 2. What does circling want? 3. Meditations on moloch 4. Memetics 9. When thinkers are way ahead of their time, how are they ahead, and how can that be replicated? 10. How to make your emotions work for you, consistently. 11. Mastery 12. Arrogance 13. Judgmentalness 14. Learning from Criticism 15. Using People 16. Control 17. The mechanics of Ego 18. Empirical Observations of my Sexuality 19. Decompose Identity (Christianity as making identity about beliefs, not grounded in reality) 20. Differences between thinking in Narrative vs. thinking Conceptually / Analytically / Statistically 21. Meaning Maximization 1. Oh my god. How have I not created this yet? It its most beautiful form? 22. Groundedness / Ungroundedness 1. Refers to levels of analysis, map / territory conflict 23. Self-awareness 1. What is self-awareness? 2. How to maximize self-awareness? 24. Why ranking is the most dark thing that happens all the time (almost necessarily, tacitly) 1. Invalidation of specialness narratives at scale. Better than, worse than, not enough. Creates deep feelings of inadequacy. 25. How not to get trapped in a thought loop 1. How to see thought loops 26. Deconstruct ‘Values’ 1. Intuition pumps, different definitions, connotations, contexts in which it varies, meta-values, etc. 27. Meta-Identity 28. Question Generators 29. Meta-Generation 30. Superforecasting 31. To Create: Fundamental Questions in Representation Learning 32. I want to create a curriculum for representation learning that’s OLD. Exclusively papers from 2000 or earlier. (Hint: Old papers only cite other old papers. And check ICA.) 33. Take every objective in ‘What Makes a Representation Good’, add my own objectives, and for each one specify: 1. A way (or set of ways) to measure the objective 1. Distinguish between the concept of the objective and the mathematical instantiation of the objective (unless they’re truly identical) 2. The downstream consequences of doing better or worse on the objective 3. Compare two different networks over the objective 4. The rationale (and intuition pumps) for the objective 1. The counterarguments 34. Apply inversion to all of the ideas in a frontline paper in representation learning. When it works well, you’ve discovered something you think is true that others disagree with. And if it’s a foundational assumption, you can get started on making progress. 35. How do we know what we claim to know in Representation learning? Ask this of a shortlist of ‘sacred beliefs’. 36. Create a ‘sacred beliefs’ in representation learning list. 37. Create a ‘consistently questioned beliefs’ in representation learning list. 38. Why learn Discrete / Sparse Representations? Be able to give a fully fledged, fully throated defence and attack. 39. Why ‘representation’ is this crazily important concept. The dramatic, windfall differences that come out of slightly different representations. 40. Survey all possible papers I could push hard at in Abstract Representation Learning 1. Explicate all my categories of idea as low level ideas 2. Generate new categories of idea 3. List out all of the goals for representation learning as a field and multiple pathways that would fulfill each goal 1. Order the goals in terms of importance 4. List out the unknowns, the missing categories, the assumptions behind the goals, and the mistakes 5. List of likely to be true / likely to be false assumptions, and ways to prove or disprove each assumption 41. Transfer between each related field and representation learning 42. Decide what the goal is. Work backwards to research paths that accomplish the goal. Value parts of the research frontier insofar as they relate to the goal. 43. What would a benevolent authoritarian do with governmental control? 1. Creation of umpteen programs creating citizens with valuable skillsets. 44. Believability is such an important concept. 1. My own lack of it is destroying me. It’s destroying my goal setting, destroying my ability to execute. 2. It’s why people don’t trust x: she’s not believable. 3. It’s what I’ve been trying to build with reputation, but it’s not equivalent to reputation. Reputation can be fake. Believability is real. 4. It’s about the state of your models and how they interact with reality. 5. It’s about whether what you say will happen happens. 45. What is generalization? Is there a right answer to how to operate “out of domain?” 46. Map out existing value systems 47. Holes / weaknesses in existing value systems 48. This visualization technique (first walking through the motions of the necessary action in your mind, and then treat real life as just another vivid visualization) is extremely powerful and symmetric. 1. Apply sys creativity to it. And use it, over and over again. 49. Arguments have statistical properties. Map out the properties of common forms of argument. 1. For example, person1 won’t take action1, he belongs to category1. (Implicit, other agents in category1 tend not to / can’t / won’t take action1, this is an example of generalization) One could treat this as an inductive prediction and use the 50. Emotional Transmogrification 1. Redirection 51. Map out emotional space 1. Language patterns / body language for each emotion 52. Systematizing systematizing 53. Planning 54. Optimal Sleep 1. Falling asleep 2. Setting up for sleep 3. Sleeping deeply 4. Dreaming 1. Lucidity 2. As creativity 3. As problem solving 5. Getting to sleep on time 6. Getting out of bed upon waking 55. AI Memespace 1. Emotional foundations to all of the memes that propagate amongst ML researchers 56. Cult Leadership Playbook 57. Deconstructing Incentive Structures 1. Reputation systems 1. Equilibria in Uber / Lyft ratings 2. Lack of incentives to rate 3. Use of gains in reputation to motivate 1. Open source contribution 2. Sharing in social networks 3. Publishing 2. Market Mechanism 3. Blockchain 4. Bitcoin Mining 5. Skin in the Game 1. This is about creating downside risk - it’s fear of loss, not hope for gain. 6. Auctions 7. Prediction Markets 8. Peer-to-Peer 9. Credit Assignment in Research 1. Citation 2. Other forms of credit 10. Insurance 1. Education as insurance 2. Abstract and generalize - what is actually about a sense of security / safety, that people say is about something else? 1. Salary 11. Psychological interactions with all of these mechanisms 1. Use of fear in reputation - fear of loss 12. Invert 1. Fear of loss vs. gains 2. Reward vs. Punishment 13. Convenience 1. Underrated, extremely important 2. Easiest to perform action as Default 3. Abstraction as source of money - create a nice interface to harsh reality 14. Organizational incentive structures, how to get promotion, how to get status, etc. 1. This literature exists. And it’s evolved over centuries. 15. The impact of measurability on all of these 1. In reputation systems, other people’s opinions are competence 2. Distinction between mechanisms in reality vs. social reality. 1. Reality is skin in the game, often. Social reality is a layer over reality that often destroys skin in the game 2. Social reality has evolved, itself. It’s cooperation. But it has to stay at a lower level (individuals), not a higher level (firms) 16. Scalable vs. Non-Scalable Mechanisms --- *Source: [Original Google Doc](https://docs.google.com/document/d/15tU8xm_RyGbEIOraUVwhCgGN2fPr0nKCrokV395QIH8/edit?usp=sharing&sa=D&ust=1596495076361000&usg=AOvVaw0kQsxqkVMesOB3WqdLHfpQ)*