LessWrong 2.0 Reader

I asked for further details on the 10th point and Claude listed a bunch of stuff I’ve absolutely never heard of. I’d say it’s probably related to meditation if I had to guess. Here’s that.

—

Claude Let me break down Time Perception Management into its deeper components, as this is one of the most subtle yet powerful micro-skills...

1. Understanding Psychological Time vs Clock Time

• How different mental states alter time perception
  • Flow states can make hours feel like minutes
  • Anxiety can make minutes feel like hours
• Learning to intentionally shift between time perspectives
  • When to use event-based time vs clock-based time
  • How to enter "timeless" states for creative work
• Reading your internal clock
  • Understanding your natural rhythms
  • Recognizing when time perception is distorted by emotions

2. Managing Time Horizons

• Developing multiple concurrent time frames
  • Thinking in day/month/year/decade simultaneously
  • Balancing immediate tasks with long-term vision
• Time horizon switching
  • When to zoom out to years
  • When to zoom in to minutes
  • How to maintain strategic clarity while handling tactics
• Future-self continuity
  • Techniques for making future consequences feel real
  • Building strong mental connections to your future self

3. Chronotype Optimization

• Understanding your natural energy patterns
  • Peak creative hours
  • Peak analytical hours
  • Recovery periods
• Designing your life around your chronotype
  • When to schedule different types of work
  • How to handle obligations that fight your natural rhythm
• Seasonal adjustments
  • Adapting to changing daylight
  • Modifying routines for different seasons

4. Timeline Compression and Expansion Techniques

• How to make time feel longer when needed
  • Breaking experiences into smaller chunks
  • Increasing novelty to slow perceived time
  • Using attention to expand moments
• How to make time feel shorter when beneficial
  • Chunking similar tasks
  • Creating flow states
  • Using routine to speed through necessary tasks

5. Personal Tempo Awareness

• Understanding your natural pace
  • Speaking tempo
  • Walking pace
  • Decision-making speed
• Tempo matching and mismatching
  • When to match others' tempo
  • When to intentionally mismatch
  • How to set the tempo in group situations
• Rhythm management
  • Creating sustainable work rhythms
  • Building in recovery cycles
  • Recognizing when to speed up or slow down

6. Advanced Applications:

• Using time perception in negotiations
  • Creating or relieving time pressure
  • Understanding how others perceive time
• Memory engineering
  • Making good times feel longer in memory
  • Making difficult times feel shorter in memory
• Social time management
  • Managing group time perception
  • Creating shared time experiences
• Decision timing
  • Understanding when to decide quickly vs slowly
  • Recognizing optimal timing windows

This skill set is particularly valuable because most people are unconscious of how their time perception affects their decisions and experiences. Mastering it gives you significant advantages in:

• Personal productivity
• Emotional regulation
• Strategic planning
• Social influence
• Life satisfaction

Would you like me to elaborate on any particular aspect? Or shall we explore how this connects with some of the other micro-skills we discussed?​​​​​​​​​​​​​​​​

Your reaction seems fair, thanks for your thoughts! Its a good a suggestion to add an epistemic status - I'll be sure to add one next time I write something like this.

Do you think putting extra effort into learning about existing empirical work while doing conceptual work would be sufficient for good conceptual work or do you think people need to be producing empirical work themselves to really make progress conceptually?

I recently saw a tweet that asked GPT, what micro skills a person should master. I asked Claude a similar question, though had preloaded the conversation with “things a modern day Aristotle would teach a 24yo student”.

I think this is probably a valuable exercise to undertake yourself. I’m providing my prompt, and Claude’s response as an ‘appetiser’ to get you interested in this task.

—

Prompt Wonderful. Now Let's think less utilized by society. What are some more niche micro skills that you would want a person who you are tutoring to have. And this is given that you want them to be as successful as possible.​​​​​​​​​​​​​​​​

—

Claude Let me think deeply about valuable micro skills that are often overlooked but can create significant advantages in life...

1. Cold Reading & Behavioral Analysis

• Understanding micro-expressions
• Picking up on subtle vocal tone changes
• Recognizing behavioral patterns in groups
• Reading body language clusters rather than isolated gestures
• This isn't about manipulation, but rather developing deeper empathy and understanding

2. Strategic Silence & Information Management

• Knowing when not to speak
• Understanding information asymmetry
• The power of selective disclosure
• Using thoughtful pauses in conversation
• The art of asking rather than telling

3. Scenario Planning & Secondary Consequence Analysis

• Ability to see 3-4 moves ahead in decisions
• Understanding cascade effects
• Identifying hidden risks and opportunities
• Planning for both success and failure scenarios
• Recognition of system dynamics

4. Social Capital Accounting

• Understanding relationship networks
• Tracking favors and obligations implicitly
• Building genuine reciprocity
• Recognition of social momentum
• Understanding status games without playing them

5. Energy Management & Recovery Tactics

• Understanding personal energy patterns
• Quick recovery techniques between intense tasks
• Environmental design for optimal performance
• Micro-rest optimization
• State switching abilities

6. Pattern Recognition Across Domains

• Transferring insights between unrelated fields
• Spotting emerging trends early
• Recognition of historical rhymes
• Understanding cyclic behaviors
• Identifying underlying principles

7. Meta-Learning Optimization

• Learning how you specifically learn best
• Creating personal knowledge management systems
• Information synthesis techniques
• Retention optimization
• Skill transfer mechanisms

8. Probabilistic Thinking & Calibration

• Understanding base rates
• Recognizing cognitive biases in real-time
• Bayesian updating in daily life
• Risk assessment calibration
• Decision journaling

9. Attention Direction & Management

• Understanding attention as a finite resource
• Strategic ignorance
• Focus maintenance techniques
• Context switching optimization
• Signal-to-noise ratio improvement

10. Time Perception Management

• Understanding psychological time vs clock time
• Managing time horizons
• Chronotype optimization
• Timeline compression and expansion techniques
• Personal tempo awareness

Would you like me to elaborate on any of these? I find them fascinating because they're rarely taught explicitly yet can be transformative when mastered. I'm curious which ones resonate most with you?

These skills compound over time and create second-order advantages that aren't immediately obvious to others. They're also relatively timeless - they worked in ancient Athens and still work today.​​​​​​​​​​​​​​​​

you assert that you know what the monopoly spends.

First, don't we know that? It's a public company and it has to report what it spends.

But more importantly, I do generally think getting a regulated monopoly like this to become more efficient is intractable, at least in the short to medium term.

Glad to see public writing on what safety cases could look like!

I liked the section "Common precautions". Points like "wherever possible, run supervised fine-tuning on the eval", "run the eval on multiple model sizes", "ensure that we don’t train against measurements" and "all generated outputs are monitored" are commonly acknowledged as useful/important (even if obvious), so it's nice they were explicitly mentioned here.

In the AI control sketch, there's a part that I find unsatisfactory:

The situations for the game were generated using Simulation Procedure S, which we are confident generates situations from a distribution that the model cannot distinguish from the genuine deployment distribution. (See below for validating the indistinguishability of this distribution.)

It's a strong assumption that we have such a simulation procedure, at least if we are talking about General External Deployment (GED). In the Limitations section later, you write

Diversity of simulations. It seems very difficult to produce realistic simulations for a sufficiently diverse set of situations to achieve indistinguishability. Manual construction is infeasible (note that context lengths are likely to be very long). Automatically generating realistic contexts is a challenging research problem.

but I think that even this is still understating the issue: For example, during GED a user might ask a model to proofread a yet-unpublished graduate level textbook on analytic number theory she has been writing. It would be very costly to generate a diverse set of such inputs for the evaluation, and automatic generation faces the challenge of evaluation being easier than generation.

The best solution for this that I'm aware of is to train and use a trusted model to flag inputs that likely weren't in the evaluation set. Now, you mention a related technique in "[P2.5] Optional Steps."

Information hiding. Each input to an oversight model is first fed to a smaller, trusted model trained to check for clear signs that an input is either test or deployment.  Such details (such as clear evidence of the date, or other side channels) are explicitly censored in all inputs such that the presence of censorship does not itself carry information about deployment status.

but I think this is far from enough: in the "textbook on analytic number theory" example, it's extremely hard to hide the fact that this input would have been costly to simulate.^[1]

My answer would be to have a human oversee the model's actions on inputs flagged by the trusted model, and use such inputs to continually train and evaluate the model. But this is costly and not privacy-preserving. (The cost is acceptable if these inputs are rare enough, though. And if such inputs are common, one needs to put more effort to bridge the evaluation-deployment gap.)

1. ^{^}

   I also worry about multimodality (e.g. native computer usage) making it harder to 1) make realistic simulations and 2) modify inputs in real time to hide information.

Got it, that makes sense. I think I was trying to get at something like this when I was talking about constraints/selection pressure (a system has less need to use abstractions if its compute is unconstrained or there is no selection pressure in the 'produce short/quick programs' direction) but your explanation makes this clearer. Thanks again for clearing this up!

We’ll refer to these as “Bookkeeping Rules”, since they feel pretty minor if you’re already comfortable working with Bayes nets. Some examples:

• We can always add an arrow to a diagram (assuming it doesn’t introduce a loop), and the approximation will get no worse.

Here's something that's kept bothering me on and off for the last few months: This graphical rule immediately breaks Markov equivalence. Specifically, two DAGs are Markov-equivalent only if they share an (undirected) skeleton. (Lemma 6.1 at the link.)

If the major/only thing we care about here regarding latential Bayes nets is that our Grand Joint Distribution $P\left[X_G\right]$ factorize over (that is, satisfy) our DAG $G$ (and all of the DAGs we can get from it by applying the rules here), then by Thm 6.2 in the link above, $P$ is also globally/locally Markov wrt $G$. This holds even when $P\left[X_G\right]>0$ is not guaranteed for some of the possible joint states in $X_G$, unlike Hammersley-Clifford would require.

That in turn means that (Def 6.5) there's some distributions $P$ can be such that $P\left[X_G\right]$ factors over $G$, but not $G^{\prime }=G+anonloopyextraarrow$ (where $G^{\prime }$ trivially has the same vertices as $G$ does); specifically, because $G,G^{\prime }$ don't (quite) share a skeleton, they can't be Markov-equivalent, and because they aren't Markov-equivalent, $P$ no longer needs to be (locally/globally) Markov wrt $G^{\prime }$ (and in fact there must exist some $P$ which explicitly break this), and because of that, such $P$ need not factor over $G^{\prime }$. Which I claim we should not want here, because (as always) we care primarily about preserving which joint probability distributions factorize over/satisfy which DAGs, and of course we probably don't get to pick whether our $P$ is one of the ones where that break in the chain of logic matters.

This is a little like the case of the Haruhi Problem, where a significant advance regarding the number of superpermutations was made by an anonymous poster on 4chan. In that case, the ephemerality of the post was a reasonable concern, and the solution was someone reposted the proof on ArXiv OEIS (with "Anonymous 4chan Poster" as the first author), and then cited that.

Here, you have a fixed url, so you could just follow the established conventions for citing webpages. I don't think you need any special justification for it, nor do you need to treat this as anything other than "merely another online website" (you don't think it's "snobbish or discriminatory" to pretend it's something more because you count yourself among its users?). 

I didn’t read it very carefully but how would you respond to the dilemma:

• If the programmer has to write things like “tires are black” into the source code, then it’s totally impractical. (…pace davidad & Doug Lenat.)
• If the programmer doesn’t have to write things like “tires are black” into the source code, then presumably a learning algorithm is figuring out things like “tires are black” from unlabeled data. And then you’re going to wind up with some giant data structure full of things like “ENTITY 92852384 implies ENTITY 8593483 with probability 0.36”. And then we have an alignment problem because the AI’s goals will be defined in terms of these unlabeled entities [LW · GW] which are hard to interpret, and where it’s hard to guess how they’ll generalize after reflection, distributional shifts, etc.

I’m guessing you’re in the second bullet but I’m not sure how you’re thinking about this alignment concern.