Just Saying What You Mean Is Impossible by Zvi Moshowitz—“Humans are automatically doing lightning fast implicit probabilistic analysis on social information in the background of every moment of their lives.” This implies there is no way to divorce the content of your communication from its myriad probabilistic social implications. Different phrasings will just send different implications.
In Defense Of Individualist Culture by Sarah Constantin (Otium) - A description of individualist culture. Criticisms of individualist culture: Lacking sympathy, few good defaults. Defenses: Its very hard to change people (psychology research review). A defense of naive personal identity. Traditional culture is fragile. Building a community project is hard in the modern world, prepare for the failure modes. Modernity has big upsides, some people will make better choices than the traditional rules allow.
My Current Thoughts On Miris Highly Reliable by Daniel Dewey (EA forum) - Report by the Open Phil AI safety lead. A basic description of and case for the MIRI program. Conclusion: 10% credence in MIRI’s work being highly useful. Reasons: Hard to apply to early agents, few researchers are excited, other approaches seem more promising.
Conversation With Dario Amodei by Jeff Kaufman—“The research that’s most valuable from an AI safety perspective also has substantial value from the perspective of solving problems today”. Prioritize work on goals. Transparency and adversarial examples are also important.
Cfar Week 1 by mindlevelup—What is working at CFAF actually like. Less rationality research than anticipated. Communication costs scale quadratically. Organization efficiency and group rationality.
The Ladder Of Interventions by mindlevelup—“This is a hierarchy of techniques to use for in-the-moment situations where you need to “convince” yourself to do something.” The author uses these methods in practice.
On Dragon Army by Zvi Moshowitz—Long response to many quotes from “Dragon Army Barracks”. Duncan’t attitude to criticism. Tyler Durden shouldn’t appeal to Duncan. Authoritarian group houses haven’t been tried. Rationalists undervalue exploration. Loneliness and doing big things. The pendulum model of social progress. Sticking to commitments even when its painful. Saving face when you screw up. True Reliability: The bay is way too unreliable but Duncan goes too far. Trust and power Dynamics. Pragmatic criticism of the charter.
Without Belief In A God But Never Without Belief In A Devil by Lou (sam[]zdat) - The nature of mass movements. The beats and the John Birchers. The taxonomy of the frustrated. Horseshoe theory. The frustrated cannot derive satisfaction from action, something else has to fill the void Poverty, work and meaning. Mass movements need to sow resentment. Hatred is the strongest unifier. Modernity inevitably causes justified resentment. Tocqueville, Polyanai, Hoffer and Scott’s theories. Helpful and unhelpful responses.
On The Effects Of Inequality On Economic Growth by Artir (Nintil) - Most of the article tries to explain and analyze the economic consensus on whether inequality harms growth. A very large number of papers are cited and discussed. A conclusion that the effect is at most small.
===Scott:
Two Kinds Of Caution by Scott Alexander—Sometimes boring technologies (ex container ships) wind up being far more important than flashy tech. However Scott argues that often the flashy tech really is important. There is too much contrarianism and not enough meta-contrarianism. AI risk.
Why Are Transgender People Immune To Optical Illusions by Scott Alexander—Scott’s community survey showed, with a huge effect size, that transgender individuals are less susceptible to the spinning mask and dancer illusions. Trans suffer from dis-associative disorder at a high rate. Connections between the two phenomena and NDMA. Commentary on the study methodology.
Rethinking Reality And Rationality by mindlevelup—Productivity is almost a solved problem. Much current rationalist research is very esoteric. Finally grokking effective altruism. Getting people good enough at rationality that they are self correcting. Pedagogy and making research fields legible.
The Power Of Pettiness by Sarah Perry (ribbonfarm) - “These emotions – pettiness and shame – are the engines driving epistemic progress” Four virtues: Loneliness, ignorance, pettiness and overconfidence.
The Abyss Of Want by AellaGirl—The infinite regress of ‘Asking why’. Taking acid and ego death. You can’t imagine the experience of death. Coming back to life. Wanting to want things. Humility and fake enlightenment.
Epistemic Laws Of Motion by SquirrelInHell—Newton’s three laws re-interpreted in terms of psychology and people’s strategies. A worked example using ‘physics’ to determine if someone will change their mind. Short and clever.
Fictional Body Language by Eukaryote—Body language in literature is often very extreme compared to real life. Emojis don’t easily map to irl body language. A ‘random’ sample of how emotion in represented in American Gods, Earth and Lirael. Three strategies: Explicitly describing feelings vs describing actions vs metaphors.
Bayesian Probability Theory As Extended Logic A by ksvanhorn (lesswrong) - Cox’s theorem is often cited to support that Bayesian probability is the only valid fundamental method of plausible reasoning. A simplified guide to Cox’s theorem. The author their paper that uses weaker assumptions than Cox’s theorem. The author’s full paper and a more detailed exposition of Cox’s theorem are linked.
Ideas On A Spectrum by Elo (BearLamp) - Putting ideas like ‘selfishness’ on a spectrum. Putting yourself and others on the spectrum. People who give you advice might disagree with you about where you fall on the spectrum. Where do you actually stand?
A Post Em Era Hint by Robin Hanson—In past ages there were pairs of innovations that enabled the emulation age without changing the growth rate. Forager: Reasoning and language. Farmer: Writing and math. Industrial: Computers and Digital Communication. What will the em-age equivalents be?
Zen Koans by Elo (BearLamp) - Connections between koans and rationalist ideas. A large number of koans are included at the end of the post. Audio of the associated meetup is included.
Fermi Paradox Resolved by Tyler Cowen—Link to a presentation. Don’t just multiply point estimates. Which Drake parameters are uncertain. The Great filter is probably in the past. Lots of interesting graphs and statistics. Social norms and laws. Religion. Eusocial society.
Call To Action by Elo (BearLamp) - Culmination of a 21 article series on life improvement and getting things done. A review of the series as a whole and thoughts on moving forward.
Cfar Week 1 by mindlevelup—What is working at CFAF actually like. Less rationality research than anticipated. Communication costs scale quadratically. Organization efficiency and group rationality.
Onemagisterium Bayes by tristanm (lesswrong) - Toolbox-ism is the dominant mode of thinking today. Downsides of toolbox-ism. Desiderata that imply Bayesianism. Major problems: Assigning priors and encountering new hypothesis. Four minor problems. Why the author is still a strong Bayesianism. Strong Bayesians can still use frequentist tools. AI Risk.
Selfconscious Ideology by casebash (lesswrong) - Lesswrong has a self conscious ideology. Self conscious ideologies have major advantages even if any given self-conscious ideology is flawed.
Intellectuals As Artists by Robin Hanson—Many norms function to show off individual impressiveness: Conversations, modern songs, taking positions on diverse subjects. Much intellectualism is not optimized for status gains not finding truth.
Just Saying What You Mean Is Impossible by Zvi Moshowitz—“Humans are automatically doing lightning fast implicit probabilistic analysis on social information in the background of every moment of their lives.” This implies there is no way to divorce the content of your communication from its myriad probabilistic social implications. Different phrasings will just send different implications.
In Defense Of Individualist Culture by Sarah Constantin (Otium) - A description of individualist culture. Criticisms of individualist culture: Lacking sympathy, few good defaults. Defenses: Its very hard to change people (psychology research review). A defense of naive personal identity. Traditional culture is fragile. Building a community project is hard in the modern world, prepare for the failure modes. Modernity has big upsides, some people will make better choices than the traditional rules allow.
Forget The Maine by Robin Hanson—Monuments are not optimized for reminding people to do better. Instead they largely serve as vehicles for simplistic ideology.
The Ladder Of Interventions by mindlevelup—“This is a hierarchy of techniques to use for in-the-moment situations where you need to “convince” yourself to do something.” The author uses these methods in practice.
On Dragon Army by Zvi Moshowitz—Long response to many quotes from “Dragon Army Barracks”. Duncan’t attitude to criticism. Tyler Durden shouldn’t appeal to Duncan. Authoritarian group houses haven’t been tried. Rationalists undervalue exploration. Loneliness and doing big things. The pendulum model of social progress. Sticking to commitments even when its painful. Saving face when you screw up. True Reliability: The bay is way too unreliable but Duncan goes too far. Trust and power Dynamics. Pragmatic criticism of the charter.
Conversation With Dario Amodei by Jeff Kaufman—“The research that’s most valuable from an AI safety perspective also has substantial value from the perspective of solving problems today”. Prioritize work on goals. Transparency and adversarial examples are also important.
Why Don’t Ai Researchers Panic by Bayesian Investor—AI researchers predict a 5% chance of “extremely bad” (extinction level) events, why aren’t they panicking? Answers: They are thinking of less bad worst cases, optimism about counter-measures, risks will be easy to deal with later, three “star theories” (MIRI, Paul Christiano, GOFAI). More answers: Fatal pessimism and resignation. It would be weird to openly worry. Benefits of AI-safety measures are less than the costs. Risks are distant.
Strategic Implications Of Ai Scenarios by (EA forum) - Questions and topics: Advanced AI timelines. Hard or soft takeoff? Goal alignment? Will advanced AI act as a single entity or a distributed system? Implication for estimating the EV of donating to AI-safety. - Tobias Baumann
Tool Use Intelligence Conversation by The Foundational Research Institute—A dialogue. Comparisons between humans and chimps/lions. The value of intelligence depends on the available tools. Defining intelligence. An addendum on “general intelligence” and factors that make intelligence useful.
Self-modification As A Game Theory Problem by (lesswrong) - “If I’m right, then any good theory for cooperation between AIs will also double as a theory of stable self-modification for a single AI, and vice versa.” An article with mathematical details is linked.
Looking Into Ai Risk by Jeff Kaufman—Jeff is trying to decide if AI risk is a serious concern and whether he should consider working in the field. Jeff’s AI-risk reading list. A large comment section with interesting arguments.
My Current Thoughts On Miris Highly Reliable by Daniel Dewey (EA forum) - Report by the Open Phil AI safety lead. A basic description of and case for the MIRI program. Conclusion: 10% credence in MIRI’s work being highly useful. Reasons: Hard to apply to early agents, few researchers are excited, other approaches seem more promising.
How Can We Best Coordinate As A Community by Benn Todd (EA forum) - ‘Replaceability’ is a bad reason not to do direct work, lots of positions are very hard to fill. Comparative Advantage and division of labor. Concrete ways to boost productivity: 5 minute favours, Operations roles, Community infrastructure, Sharing knowledge and Specialization. EA Global Video is included.
Deciding Whether to Recommend Fistula Management Charities by The GiveWell Blog—“An obstetric fistula, or gynecologic fistula, is an abnormal opening between the vagina and the bladder or rectum.” Fistula management, including surgery. Open questions and uncertainty particularly around costs. Our plans to partner with IDinsight to answer these questions.
Hidden Cost Digital Convenience by Innovations for Poverty—Moving from in person to digital micro-finance can harm saving rates in developing countries. Reduction in group cohesion and visible transaction fees. Linked paper with details.
Projects People And Processes by Open Philosophy—Three approaches used by donors and decision makers: Choose from projects presented by experts, defer near-fully to trusted individuals and establishing systematic criteria. Pros and cons of each. Open Phil’s current approach.
Effective Altruism As Costly Signaling by Raemon (EA forum) - ” ‘a bunch of people saying that rich people should donate to X’ is a less credible signal than ‘a bunch of people saying X thing is important enough that they are willing to donate to it themselves.’ ”
Oops Prize by Ben Hoffman (Compass Rose) - Positive norms around admitting you were wrong. Charity Science publicly admitted they were wrong about grant writing. Did anyone organization at EA Global admit they made a costly mistake? 1K oops prize.
===Politics and Economics:
Scraps 3 Hoffer And Performance Art by Lou (sam[]zdat) - Growing out of radicalism. Either economic and family instability can cause mass movements. why the left has adopted Freud. The Left’s economic platform is popular, its cultural platform is not. Performance art: Marina Abramović′s’ ‘Rhythm 0’. Recognizing and denying your own power.
What Replaces Rights And Discourse by Feddie deBoer—Lots of current leftist discourse is dismissive of rights and open discussion. But what alternative is there? The Soviets had bad justifications and a terrible system but at least it had an explicit philosophical alternative.
Why Do You Hate Elua by H i v e w i r e d—Scott’s Elua as an Eldritch Abomination that threatens traditional culture. An extended sci-fi quote about Ra the great computer. “The forces of traditional values remembered an important fact: once you have access to the hardware, it’s over.”
Why Did Europe Lose Crusades by Noah Smith—Technological comparison between Europe and the Middle East. Political divisions on both sides. Geographic distance. Lack of motivation.
Data On Campus Free Speech Cases by Ozy (Thing of Things) - Ozy classifies a sample of the cases handled by the Foundation for Individual Rights in Education. Ozy classifies 77 cases as conservative, liberal or apolitical censorship. Conservative ideas were censored 52%, liberal 26% and apolitical 22%.
Beware The Moral Spotlight by Robin Hanson—The stated goals of government/business don’t much matter compared to the selective pressures on their leadership, don’t obsess over which sex has the worse deal overall, don’t overate the benefits of democracy and ignore higher impact changes to government.
Reply To Yudkowsky by Bryan Caplan—Caplan quotes and replies to many sections Yudkowsky’s response. Points: Yudkowsky’s theory is a special case of Caplan’s. The left has myriad complaints about markets. Empirically the market actually has consistently provided large benefits in many countries and times.
Without Belief In A God But Never Without Belief In A Devil by Lou (sam[]zdat) - The nature of mass movements. The beats and the John Birchers. The taxonomy of the frustrated. Horseshoe theory. The frustrated cannot derive satisfaction from action, something else has to fill the void Poverty, work and meaning. Mass movements need to sow resentment. Hatred is the strongest unifier. Modernity inevitably causes justified resentment. Tocqueville, Polyanai, Hoffer and Scott’s theories. Helpful and unhelpful responses.
On The Effects Of Inequality On Economic Growth by Artir (Nintil) - Most of the article tries to explain and analyze the economic consensus on whether inequality harms growth. A very large number of papers are cited and discussed. A conclusion that the effect is at most small.
Sexualtaboos by AellaGirl—A graph of sexual fetishes. The axes are “taboo-ness” and “reported interest”. Taboo correlated negatively with interest (p < 0.01). Lots of fetishes are included and the sample size is pretty large.
If You’re In School Try The Curriculum by Freddie deBoer—Ironic detachment “leaves you with the burden of the work but without the emotional support of genuine resolve”. Don’t be the sort of person who tweets hundreds of thousands of times but pretends they don’t care about online.
Media Recommendations by Sailor Vulcan (BYS) - Various Reviews including: Games, Animated TV shows, Rationalist Pokemon. A more detailed review of Harry Potter and the Methods of Rationality.
Sunday Assorted Links by Tyler Cowen—Variety of Topics. Ethereum Cryptocurrency, NYC Diner decline, Building Chinese Airports, Soccer Images, Drone Wars, Harberger Taxation, Douthat on Heathcare.
Rescuing The Extropy Magazine Archives by deku_shrub (lesswrong) - “You’ll find some really interesting very early articles on neural augmentation, transhumanism, libertarianism, AI (featuring Eliezer), radical economics (featuring Robin Hanson of course) and even decentralized payment systems.”
Epistemic Spot Check A Guide To Better Movement Todd Hargrove by Aceso Under Glass—Flexibility and Chronic Pain. Early section on flexibility fails check badly. Section on psychosomatic pain does much better. Model: Simplicity (Good), Explanation (Fantastic), Explicit Predictions (Good), Useful Predictions (Poor), Acknowledge Limits (Poor), Measurability (Poor).
Book Review Barriers by Eukaryote—Even cell culturing is surprisingly hard if you don’t know the details. There is not much institutional knowledge left in the field of bioweapons. Forcing labs underground makes bioterrorism even harder. However synthetic biology might make things much more dangerous.
Kyle Maynard Without Limits by Tim Ferriss—“Kyle Maynard is a motivational speaker, bestselling author, entrepreneur, and ESPY award-winning mixed martial arts athlete, known for becoming the first quadruple amputee to reach the summit of Mount Kilimanjaro and Mount Aconcagua without the aid of prosthetics.”
Myers Briggs, Diet, Mistakes And Immortality by Tim Ferriss—Ask me anything podcast. Topics beyond the title: Questions to prompt introspection, being a Jack of All Trades, balancing future and present goals, don’t follow your passion, 80⁄20 memory retention, advice to your past selves.
Avic Roy by The Ezra Klein Show—Better Care Reconciliation Act, broader health care philosophies that fracture the right. Roy’s disagreements with the CBO’s methodology. The many ways he thinks the Senate bill needs to improve. How the GOP has moved left on health care policy. Medicaid, welfare reform, and the needy who are hard to help. The American health care system subsidizes the rich, etc.
Chris Blattman 2 by EconTalk—“Whether it’s better to give poor Africans cash or chickens and the role of experiments in helping us figure out the answer. Along the way he discusses the importance of growth vs. smaller interventions and the state of development economics.”
Blake Mycoskie by Tim Ferriss—Early entrepreneurial ventures. The power of journaling. How “the stool analogy” changed Blake’s life. Lessons from Ben Franklin.
Ben Sasse by Tyler Cowen—“Kansas vs. Nebraska, famous Nebraskans, Chaucer and Luther, unicameral legislatures, the decline of small towns, Ben’s prize-winning Yale Ph.d thesis on the origins of conservatism, what he learned as a university president, Stephen Curry, Chevy Chase, Margaret Chase Smith”
Danah Boyd on why Fake News is so Easy to Believe by The Ezra Klein Show—Fake news, digital white flight, how an anthropologist studies social media, machine learning algorithms reflect our prejudices rather than fixing them, what Netflix initially got wrong about their recommendations engine, the value of pretending your audience is only six people, the early utopian visions of the internet.
Bi-Weekly Rational Feed
===Highly Recommended Articles:
Just Saying What You Mean Is Impossible by Zvi Moshowitz—“Humans are automatically doing lightning fast implicit probabilistic analysis on social information in the background of every moment of their lives.” This implies there is no way to divorce the content of your communication from its myriad probabilistic social implications. Different phrasings will just send different implications.
In Defense Of Individualist Culture by Sarah Constantin (Otium) - A description of individualist culture. Criticisms of individualist culture: Lacking sympathy, few good defaults. Defenses: Its very hard to change people (psychology research review). A defense of naive personal identity. Traditional culture is fragile. Building a community project is hard in the modern world, prepare for the failure modes. Modernity has big upsides, some people will make better choices than the traditional rules allow.
My Current Thoughts On Miris Highly Reliable by Daniel Dewey (EA forum) - Report by the Open Phil AI safety lead. A basic description of and case for the MIRI program. Conclusion: 10% credence in MIRI’s work being highly useful. Reasons: Hard to apply to early agents, few researchers are excited, other approaches seem more promising.
Conversation With Dario Amodei by Jeff Kaufman—“The research that’s most valuable from an AI safety perspective also has substantial value from the perspective of solving problems today”. Prioritize work on goals. Transparency and adversarial examples are also important.
Cfar Week 1 by mindlevelup—What is working at CFAF actually like. Less rationality research than anticipated. Communication costs scale quadratically. Organization efficiency and group rationality.
The Ladder Of Interventions by mindlevelup—“This is a hierarchy of techniques to use for in-the-moment situations where you need to “convince” yourself to do something.” The author uses these methods in practice.
On Dragon Army by Zvi Moshowitz—Long response to many quotes from “Dragon Army Barracks”. Duncan’t attitude to criticism. Tyler Durden shouldn’t appeal to Duncan. Authoritarian group houses haven’t been tried. Rationalists undervalue exploration. Loneliness and doing big things. The pendulum model of social progress. Sticking to commitments even when its painful. Saving face when you screw up. True Reliability: The bay is way too unreliable but Duncan goes too far. Trust and power Dynamics. Pragmatic criticism of the charter.
Without Belief In A God But Never Without Belief In A Devil by Lou (sam[]zdat) - The nature of mass movements. The beats and the John Birchers. The taxonomy of the frustrated. Horseshoe theory. The frustrated cannot derive satisfaction from action, something else has to fill the void Poverty, work and meaning. Mass movements need to sow resentment. Hatred is the strongest unifier. Modernity inevitably causes justified resentment. Tocqueville, Polyanai, Hoffer and Scott’s theories. Helpful and unhelpful responses.
On The Effects Of Inequality On Economic Growth by Artir (Nintil) - Most of the article tries to explain and analyze the economic consensus on whether inequality harms growth. A very large number of papers are cited and discussed. A conclusion that the effect is at most small.
===Scott:
Two Kinds Of Caution by Scott Alexander—Sometimes boring technologies (ex container ships) wind up being far more important than flashy tech. However Scott argues that often the flashy tech really is important. There is too much contrarianism and not enough meta-contrarianism. AI risk.
Open Road by Scott Alexander—Bi-weekly public open thread. Some messages from Scott Alexander.
To The Great City by Scott Alexander—Scott’s Karass is in San Fransisco. He is going home.
Open Thread 78 75 by Scott Alexander—Bi-weekly public open thread.
Why Are Transgender People Immune To Optical Illusions by Scott Alexander—Scott’s community survey showed, with a huge effect size, that transgender individuals are less susceptible to the spinning mask and dancer illusions. Trans suffer from dis-associative disorder at a high rate. Connections between the two phenomena and NDMA. Commentary on the study methodology.
Contra Otium On Individualism by Scott Alexander (Scratchpad) - Eight point summary of Sarah’s defense of individualism. Scott is terrified the market place of ideals doesn’t work and his own values aren’t memetically fit.
Conversation Deliberately Skirts The Border Of Incomprehensibility by Scott Alexander—Communication is often designed to be confusing so as to preserve plausible deniability.
===Rationalist:
Rethinking Reality And Rationality by mindlevelup—Productivity is almost a solved problem. Much current rationalist research is very esoteric. Finally grokking effective altruism. Getting people good enough at rationality that they are self correcting. Pedagogy and making research fields legible.
The Power Of Pettiness by Sarah Perry (ribbonfarm) - “These emotions – pettiness and shame – are the engines driving epistemic progress” Four virtues: Loneliness, ignorance, pettiness and overconfidence.
Irrationality is in the Eye of the Beholder by João Eira (Lettuce be Cereal) - Is eating a chocolate croissant on a diet always irrational? Context, hidden motivations and the curse of knowledge.
The Abyss Of Want by AellaGirl—The infinite regress of ‘Asking why’. Taking acid and ego death. You can’t imagine the experience of death. Coming back to life. Wanting to want things. Humility and fake enlightenment.
Epistemic Laws Of Motion by SquirrelInHell—Newton’s three laws re-interpreted in terms of psychology and people’s strategies. A worked example using ‘physics’ to determine if someone will change their mind. Short and clever.
Against Lone Wolf Selfimprovement by cousin_it (lesswrong) - Lone wolf improvement is hard. Too many rationalists attempt it for cultural and historical reasons. Its often better to take a class or find a group.
Fictional Body Language by Eukaryote—Body language in literature is often very extreme compared to real life. Emojis don’t easily map to irl body language. A ‘random’ sample of how emotion in represented in American Gods, Earth and Lirael. Three strategies: Explicitly describing feelings vs describing actions vs metaphors.
Bayesian Probability Theory As Extended Logic A by ksvanhorn (lesswrong) - Cox’s theorem is often cited to support that Bayesian probability is the only valid fundamental method of plausible reasoning. A simplified guide to Cox’s theorem. The author their paper that uses weaker assumptions than Cox’s theorem. The author’s full paper and a more detailed exposition of Cox’s theorem are linked.
Steelmanning The Chinese Room Argument by cousin_it (lesswrong) - A short thought experiment about consciousness and inferring knowledge from behavior.
Ideas On A Spectrum by Elo (BearLamp) - Putting ideas like ‘selfishness’ on a spectrum. Putting yourself and others on the spectrum. People who give you advice might disagree with you about where you fall on the spectrum. Where do you actually stand?
A Post Em Era Hint by Robin Hanson—In past ages there were pairs of innovations that enabled the emulation age without changing the growth rate. Forager: Reasoning and language. Farmer: Writing and math. Industrial: Computers and Digital Communication. What will the em-age equivalents be?
Zen Koans by Elo (BearLamp) - Connections between koans and rationalist ideas. A large number of koans are included at the end of the post. Audio of the associated meetup is included.
Fermi Paradox Resolved by Tyler Cowen—Link to a presentation. Don’t just multiply point estimates. Which Drake parameters are uncertain. The Great filter is probably in the past. Lots of interesting graphs and statistics. Social norms and laws. Religion. Eusocial society.
Developmental Psychology In The Age Of Ems by Gordan (Map and Territory) - Brief intro to the Age of Em. Farmer values. Robin’s approach to futurism. Psychological implications of most ems being middle aged. Em conservatism and maturity.
Call To Action by Elo (BearLamp) - Culmination of a 21 article series on life improvement and getting things done. A review of the series as a whole and thoughts on moving forward.
Cfar Week 1 by mindlevelup—What is working at CFAF actually like. Less rationality research than anticipated. Communication costs scale quadratically. Organization efficiency and group rationality.
Onemagisterium Bayes by tristanm (lesswrong) - Toolbox-ism is the dominant mode of thinking today. Downsides of toolbox-ism. Desiderata that imply Bayesianism. Major problems: Assigning priors and encountering new hypothesis. Four minor problems. Why the author is still a strong Bayesianism. Strong Bayesians can still use frequentist tools. AI Risk.
Selfconscious Ideology by casebash (lesswrong) - Lesswrong has a self conscious ideology. Self conscious ideologies have major advantages even if any given self-conscious ideology is flawed.
Intellectuals As Artists by Robin Hanson—Many norms function to show off individual impressiveness: Conversations, modern songs, taking positions on diverse subjects. Much intellectualism is not optimized for status gains not finding truth.
Just Saying What You Mean Is Impossible by Zvi Moshowitz—“Humans are automatically doing lightning fast implicit probabilistic analysis on social information in the background of every moment of their lives.” This implies there is no way to divorce the content of your communication from its myriad probabilistic social implications. Different phrasings will just send different implications.
In Defense Of Individualist Culture by Sarah Constantin (Otium) - A description of individualist culture. Criticisms of individualist culture: Lacking sympathy, few good defaults. Defenses: Its very hard to change people (psychology research review). A defense of naive personal identity. Traditional culture is fragile. Building a community project is hard in the modern world, prepare for the failure modes. Modernity has big upsides, some people will make better choices than the traditional rules allow.
Forget The Maine by Robin Hanson—Monuments are not optimized for reminding people to do better. Instead they largely serve as vehicles for simplistic ideology.
The Ladder Of Interventions by mindlevelup—“This is a hierarchy of techniques to use for in-the-moment situations where you need to “convince” yourself to do something.” The author uses these methods in practice.
On Dragon Army by Zvi Moshowitz—Long response to many quotes from “Dragon Army Barracks”. Duncan’t attitude to criticism. Tyler Durden shouldn’t appeal to Duncan. Authoritarian group houses haven’t been tried. Rationalists undervalue exploration. Loneliness and doing big things. The pendulum model of social progress. Sticking to commitments even when its painful. Saving face when you screw up. True Reliability: The bay is way too unreliable but Duncan goes too far. Trust and power Dynamics. Pragmatic criticism of the charter.
===AI:
Updates To The Research Team And A Major Donation by The MIRI Blog—MIRIr received a 1 million dollar donation. Two new full-time researchers. Two researchers leaving. Medium term financial plans.
Conversation With Dario Amodei by Jeff Kaufman—“The research that’s most valuable from an AI safety perspective also has substantial value from the perspective of solving problems today”. Prioritize work on goals. Transparency and adversarial examples are also important.
Why Don’t Ai Researchers Panic by Bayesian Investor—AI researchers predict a 5% chance of “extremely bad” (extinction level) events, why aren’t they panicking? Answers: They are thinking of less bad worst cases, optimism about counter-measures, risks will be easy to deal with later, three “star theories” (MIRI, Paul Christiano, GOFAI). More answers: Fatal pessimism and resignation. It would be weird to openly worry. Benefits of AI-safety measures are less than the costs. Risks are distant.
Strategic Implications Of Ai Scenarios by (EA forum) - Questions and topics: Advanced AI timelines. Hard or soft takeoff? Goal alignment? Will advanced AI act as a single entity or a distributed system? Implication for estimating the EV of donating to AI-safety. - Tobias Baumann
Tool Use Intelligence Conversation by The Foundational Research Institute—A dialogue. Comparisons between humans and chimps/lions. The value of intelligence depends on the available tools. Defining intelligence. An addendum on “general intelligence” and factors that make intelligence useful.
Self-modification As A Game Theory Problem by (lesswrong) - “If I’m right, then any good theory for cooperation between AIs will also double as a theory of stable self-modification for a single AI, and vice versa.” An article with mathematical details is linked.
Looking Into Ai Risk by Jeff Kaufman—Jeff is trying to decide if AI risk is a serious concern and whether he should consider working in the field. Jeff’s AI-risk reading list. A large comment section with interesting arguments.
===EA:
Ea Marketing And A Plea For Moral Inclusivity by MichaelPlant (EA forum) - EA markets itself as being about poverty reduction. Many EAs think other topics are more important (far future, AI, animal welfare, etc). The author suggests becoming both more inclusive and more openly honest.
My Current Thoughts On Miris Highly Reliable by Daniel Dewey (EA forum) - Report by the Open Phil AI safety lead. A basic description of and case for the MIRI program. Conclusion: 10% credence in MIRI’s work being highly useful. Reasons: Hard to apply to early agents, few researchers are excited, other approaches seem more promising.
How Can We Best Coordinate As A Community by Benn Todd (EA forum) - ‘Replaceability’ is a bad reason not to do direct work, lots of positions are very hard to fill. Comparative Advantage and division of labor. Concrete ways to boost productivity: 5 minute favours, Operations roles, Community infrastructure, Sharing knowledge and Specialization. EA Global Video is included.
Deciding Whether to Recommend Fistula Management Charities by The GiveWell Blog—“An obstetric fistula, or gynecologic fistula, is an abnormal opening between the vagina and the bladder or rectum.” Fistula management, including surgery. Open questions and uncertainty particularly around costs. Our plans to partner with IDinsight to answer these questions.
Allocating the Capital by GiveDirectly—Eight media links on Give Directly, Basic Income and Cash Transfers.
Testing An Ea Networkbuilding Strategy by remmelt (EA forum) - Pivot from supporting EA charities to cooperating with EA networks. Detailed goals, strategy, assumptions, metrics, collaborators and example actions.
How Long Does It Take To Research And Develop A Vaccine by (EA forum) - How long it takes to make a vaccine. Literature review. Historical data on how long a large number of vaccines took to develop. Conclusions.
Hi Im Luke Muehlhauser Ama About Open by Luke Muelhauser (EA forum) - Animal and computer consciousness. Luke wrote a report for the open philanthropy project on consciousness. Lots of high quality questions have been posted.
Hidden Cost Digital Convenience by Innovations for Poverty—Moving from in person to digital micro-finance can harm saving rates in developing countries. Reduction in group cohesion and visible transaction fees. Linked paper with details.
Projects People And Processes by Open Philosophy—Three approaches used by donors and decision makers: Choose from projects presented by experts, defer near-fully to trusted individuals and establishing systematic criteria. Pros and cons of each. Open Phil’s current approach.
Effective Altruism An Idea Repository by Onemorenickname (lesswrong) - Effective altruism is less of a closed organization than the author thought. Building a better platform for effective altruist idea sharing.
Effective Altruism As Costly Signaling by Raemon (EA forum) - ” ‘a bunch of people saying that rich people should donate to X’ is a less credible signal than ‘a bunch of people saying X thing is important enough that they are willing to donate to it themselves.’ ”
The Person Affecting Philanthropists Paradox by MichaelPlant (EA forum) - Population ethics. The value of creating more happy people as opposed to making pre-existing people happy. Application to the question of whether to donate now or invest and donate later.
Oops Prize by Ben Hoffman (Compass Rose) - Positive norms around admitting you were wrong. Charity Science publicly admitted they were wrong about grant writing. Did anyone organization at EA Global admit they made a costly mistake? 1K oops prize.
===Politics and Economics:
Scraps 3 Hoffer And Performance Art by Lou (sam[]zdat) - Growing out of radicalism. Either economic and family instability can cause mass movements. why the left has adopted Freud. The Left’s economic platform is popular, its cultural platform is not. Performance art: Marina Abramović′s’ ‘Rhythm 0’. Recognizing and denying your own power.
What Replaces Rights And Discourse by Feddie deBoer—Lots of current leftist discourse is dismissive of rights and open discussion. But what alternative is there? The Soviets had bad justifications and a terrible system but at least it had an explicit philosophical alternative.
Why Do You Hate Elua by H i v e w i r e d—Scott’s Elua as an Eldritch Abomination that threatens traditional culture. An extended sci-fi quote about Ra the great computer. “The forces of traditional values remembered an important fact: once you have access to the hardware, it’s over.”
Why Did Europe Lose Crusades by Noah Smith—Technological comparison between Europe and the Middle East. Political divisions on both sides. Geographic distance. Lack of motivation.
Econtalk On Generic Medications by Aceso Under Glass—A few egregious ways that big pharma games the patent system. Short.
Data On Campus Free Speech Cases by Ozy (Thing of Things) - Ozy classifies a sample of the cases handled by the Foundation for Individual Rights in Education. Ozy classifies 77 cases as conservative, liberal or apolitical censorship. Conservative ideas were censored 52%, liberal 26% and apolitical 22%.
Beware The Moral Spotlight by Robin Hanson—The stated goals of government/business don’t much matter compared to the selective pressures on their leadership, don’t obsess over which sex has the worse deal overall, don’t overate the benefits of democracy and ignore higher impact changes to government.
Reply To Yudkowsky by Bryan Caplan—Caplan quotes and replies to many sections Yudkowsky’s response. Points: Yudkowsky’s theory is a special case of Caplan’s. The left has myriad complaints about markets. Empirically the market actually has consistently provided large benefits in many countries and times.
Without Belief In A God But Never Without Belief In A Devil by Lou (sam[]zdat) - The nature of mass movements. The beats and the John Birchers. The taxonomy of the frustrated. Horseshoe theory. The frustrated cannot derive satisfaction from action, something else has to fill the void Poverty, work and meaning. Mass movements need to sow resentment. Hatred is the strongest unifier. Modernity inevitably causes justified resentment. Tocqueville, Polyanai, Hoffer and Scott’s theories. Helpful and unhelpful responses.
Genetic Behaviorism Supports The Influence Of Chance On Life Outcomes by Freddie deBoer—Much of the variance in many traits is non-shared-environment. Much non-shared-environment can be thought of as luck. In addition no one chooses or deserves their genes.
Yudkowsky On My Simpistic Theory of Left and Right by Bryan Caplan—Yudkowsky claims the left holds the market to the same standards as human beings. The market as a ritual holding back a dangerous Alien God. Caplan doesn’t respond he just quotes Yudkowsky.
On The Effects Of Inequality On Economic Growth by Artir (Nintil) - Most of the article tries to explain and analyze the economic consensus on whether inequality harms growth. A very large number of papers are cited and discussed. A conclusion that the effect is at most small.
===Misc:
Erisology Of Self And Will Representative Campbell Speaks by Everything Studies—An exposition of the “mainstream” view of the self and free will.
What Is The Ein Sof The Meaning Of Perfection In by arisen (lesswrong) - “Kabbalah is based on the analogy of the soul as a cup and G-d as the light that fills the cup. Ein Sof, nothing (“Ein”) can be grasped (“Sof”-limitation).”
Sexualtaboos by AellaGirl—A graph of sexual fetishes. The axes are “taboo-ness” and “reported interest”. Taboo correlated negatively with interest (p < 0.01). Lots of fetishes are included and the sample size is pretty large.
Huffman Codes Problem by protokol2020 - Find the possible Huffman Codes for all twenty-six English letters.
If You’re In School Try The Curriculum by Freddie deBoer—Ironic detachment “leaves you with the burden of the work but without the emotional support of genuine resolve”. Don’t be the sort of person who tweets hundreds of thousands of times but pretends they don’t care about online.
Media Recommendations by Sailor Vulcan (BYS) - Various Reviews including: Games, Animated TV shows, Rationalist Pokemon. A more detailed review of Harry Potter and the Methods of Rationality.
Sunday Assorted Links by Tyler Cowen—Variety of Topics. Ethereum Cryptocurrency, NYC Diner decline, Building Chinese Airports, Soccer Images, Drone Wars, Harberger Taxation, Douthat on Heathcare.
Summary Of Reading April June 2017 by Eli Bendersky—Brief reviews. Various topics: Heavy on Economics. Some politics, literature and other topics.
Rescuing The Extropy Magazine Archives by deku_shrub (lesswrong) - “You’ll find some really interesting very early articles on neural augmentation, transhumanism, libertarianism, AI (featuring Eliezer), radical economics (featuring Robin Hanson of course) and even decentralized payment systems.”
Epistemic Spot Check A Guide To Better Movement Todd Hargrove by Aceso Under Glass—Flexibility and Chronic Pain. Early section on flexibility fails check badly. Section on psychosomatic pain does much better. Model: Simplicity (Good), Explanation (Fantastic), Explicit Predictions (Good), Useful Predictions (Poor), Acknowledge Limits (Poor), Measurability (Poor).
Book Review Barriers by Eukaryote—Even cell culturing is surprisingly hard if you don’t know the details. There is not much institutional knowledge left in the field of bioweapons. Forcing labs underground makes bioterrorism even harder. However synthetic biology might make things much more dangerous.
Physics Problem 2 by protokol2020 - Can tidal forces rotate a metal wheel?
Poems by Scott Alexander (Scratchpad) - Violets aren’t blue.
Evaluating Employers As Junior Software by Particular Virtue—You need to write alot of code and get detailed feedback to improve as an engineer. Practical suggestions to ensure your first job fulfills both conditions.
===Podcast:
Kyle Maynard Without Limits by Tim Ferriss—“Kyle Maynard is a motivational speaker, bestselling author, entrepreneur, and ESPY award-winning mixed martial arts athlete, known for becoming the first quadruple amputee to reach the summit of Mount Kilimanjaro and Mount Aconcagua without the aid of prosthetics.”
85 Is This The End Of Europe by Waking Up with Sam Harris—Douglas Murray and his book ‘The Strange Death of Europe: Immigration, Identity, Islam’.
Myers Briggs, Diet, Mistakes And Immortality by Tim Ferriss—Ask me anything podcast. Topics beyond the title: Questions to prompt introspection, being a Jack of All Trades, balancing future and present goals, don’t follow your passion, 80⁄20 memory retention, advice to your past selves.
Interview Ro Khanna Regional Development by Tyler Cowen—Bloomberg Podcast. “Technology, jobs and economic lessons from his perspective as Silicon Valley’s congressman.”
Avic Roy by The Ezra Klein Show—Better Care Reconciliation Act, broader health care philosophies that fracture the right. Roy’s disagreements with the CBO’s methodology. The many ways he thinks the Senate bill needs to improve. How the GOP has moved left on health care policy. Medicaid, welfare reform, and the needy who are hard to help. The American health care system subsidizes the rich, etc.
Chris Blattman 2 by EconTalk—“Whether it’s better to give poor Africans cash or chickens and the role of experiments in helping us figure out the answer. Along the way he discusses the importance of growth vs. smaller interventions and the state of development economics.”
Landscapes Of Mind by Waking Up with Sam Harris—“why it’s so hard to predict future technology, the nature of intelligence, the ‘singularity’, artificial consciousness.”
Blake Mycoskie by Tim Ferriss—Early entrepreneurial ventures. The power of journaling. How “the stool analogy” changed Blake’s life. Lessons from Ben Franklin.
Ben Sasse by Tyler Cowen—“Kansas vs. Nebraska, famous Nebraskans, Chaucer and Luther, unicameral legislatures, the decline of small towns, Ben’s prize-winning Yale Ph.d thesis on the origins of conservatism, what he learned as a university president, Stephen Curry, Chevy Chase, Margaret Chase Smith”
Danah Boyd on why Fake News is so Easy to Believe by The Ezra Klein Show—Fake news, digital white flight, how an anthropologist studies social media, machine learning algorithms reflect our prejudices rather than fixing them, what Netflix initially got wrong about their recommendations engine, the value of pretending your audience is only six people, the early utopian visions of the internet.
Robin Feldman by EconTalk—Ways pharmaceutical companies fight generics.
Jason Weeden On Do People Vote Based On Self Interest by Rational Speaking—Do people vote based on personality, their upbringing, blind loyalty or do they follow their self interest? What does self-interest even mean?
Reid Hoffman 2 by Tim Ferriss—The 10 Commandments of Startup Success according to the extremely successful investor Reid Hoffman.