Rational Feed

Scott:

The Economic Perspective On Moral Standards by Scott Alexander – Scott tries to deal with the fact that ethical systems such as utilitarianism seem ‘infinitely demanding’. He defends a norm of trying to do better than average.

Preschool: Much More Than You Wanted To Know by Scott Alexander – ‘Good randomized controlled trials have shown that preschools do not improve test scores in a lasting way. Sometimes test scores go up a little bit, but these effects disappear after a year or two of regular schooling. However, early RCTs of intensive “wrap-around” preschools like the Perry Preschool Program and the Abecedarians found that graduates of those programs went on to have markedly better adult outcomes, including higher school graduation rates, more college attendance, less crime, and better jobs. But these studies were done in the 60s, before people invented being responsible, and had kind of haphazard randomization and followup. They were also small sample sizes, and from programs that were more intense than any of the scaled-up versions that replaced them. Modern scaled-up preschools like Head Start would love to be able to claim their mantle and boast similar results. But the only good RCT of Head Start, the HSIS study, is still in its first few years. It’s confirmed that Head Start test score gains fade out. But it hasn’t been long enough to study whether there are later effects on life outcomes. We can expect those results in ten years or so. For now, all we have is speculation based on a few quasi-experiments.’

Rationalist:

Preference Changing by Brian Lui – A five step process for changing your preferences: ‘To change a preference we must first confront its reality. To change our preferences, we need to deeply understand them. Next, we need to increase the contrast between instances of the desired behavior and the current unwanted behavior. Next, we reflect on whether our preference is a need or a want, and how that interacts with our conception of our identity. Once we’ve detached a preference from our identity, the final step is to examine why we engaged in the behavior in the first place.’

Book Post For October by Ozy – Book reviews: The Demon Haunted World, Teaching Your Child The Foundations of Classical Education, Star Wars: Thrawn, Dark Lord’s Answer.

Social Media Lessons by Robin Hanson – How the internet has changed now that ‘ordinary people’ instead of nerds are the primary users. Ordinary people have systematically different preferences. For example they care much less about privacy and much more about gossip.

World Government Risks Collective Suicide by Robin Hanson – If decisions are made by a single entity then it becomes very hard to avoid suicidal decisions over long time horizons. however if many decision makers agree than its much easier to avoid ruinous decisions long term. This fact makes ‘world government’ very risky.

Non Conformist Influence by Robin Hanson – A mathematical model that implies non-conformists have outsized influence. [brief]

How To Fund Prestige Science by Robin Hanson – Using prediction markets to rate scientific papers. The main idea is that in 30 years a committee will review the paper you are betting on. Hopefully 30 years is enough time for its impact, or lack thereof, to become unambiguous.

AI:

Alignment Newsletter 32 by rohinmshah – Weekly AI newsletter. OpenAI released a guide to getting started with reinforcement learning.

Podcast:

Is Philanthropy Bad For Democracy by Rationally Speaking – ‘Rob and Julia debate his criticisms of philanthropy: Does it deserve to be tax-deductible? Is it a violation of the autonomy of recipients to attach strings to their charitable gifts? And do philanthropists have too much power in society?’

Addiction, Depression And A Meaningful Life by Waking Up with Sam Harris – Sam Harris speaks with Johann Hari about his books “Chasing the Scream” and “Lost Connections.”

Advertisements

Rational Feed

Scott:

SSRIs An Update by Scott Alexander – Even treatments that are definitely effective have ‘small’ effect sizes of 0.3 to 0.5. Scott’s explanation is that many treatments work well on a subset of people but the majority are non-responders. The same is probably true for SSRIs.

Rationalist:

Mandatory Obsessions by Jacob Falkovich – Diversity of both opinions and levels of caring about political topics is valuable. However many people feel compelled to stamp it out. Why? Two explanations: They are unconsciously trying to increase their own status, they are infected by a mind virus.

Avoiding Blame By Preventing Life by Robin Hanson – What sort of robots should we build if robot workers will replace humans.

Burnout: What Is It And How To Treat It by Aceso Under Glass – What is burnout and how to treat it. TLDR: Social support and sleep are good. Ambiguity is bad. Vacations are meh.

Politics and Economics:

Mars by Robin Hanson – Terraforming Mars and when to trust for profit firms.

Podcast:

Andres From The Qualia Research Institute by The Bayesian Conspiracy – What is qualia and why can we investigate it scientifically.

Rational Feed

Scott:

Preschool: I Was Wrong by Scott Alexander – Scott admits that pre-school does have long term benefits, at least for the poor. However it seems like the benefits are not academic!

Rationalists Shouldn’t Win by Scott Alexander – Scott argues that it was mistake for rationalists to focus on individual winning. Self improve and abstract epistemic rationality require different mindsets.

Marijuana: An Update by Scott Alexander – Marijuana legalization probably increases car accidents at least 5-6%. Crashes might increase more as pot becomes further normalized. The downsides of car crashes make Scott skeptical the net effect of legalization is positive.

My California Ballot by Scott Alexander – How Scott plans to vote in California.

Penelopen Thread by Scott Alexander – Biweekly public open thread. Comment of the week is on ‘Why most of what you read online is written by crazy people’.

AI:

Alignment Newsletter 31 by rohinmshah – Weekly AI newsletter. Discussion of an improved implementation of curiosity based learning.

Learning Concepts With Energy Functions by Open Ai – “We’ve developed an energy-based model that can quickly learn to identify and generate instances of concepts, such as near, above, between, closest, and furthest, expressed as sets of 2d points. Our model learns these concepts after only five demonstrations. We also show cross-domain transfer: we use concepts learned in a 2d particle environment to solve tasks on a 3-dimensional physics-based robot.”

Podcast:

Eric Schmidt by Conversations with Tyler – “Tyler questioned Schmidt about underused management strategies, what Google learned after interviewing one job candidate sixteen times, his opinion on early vs. late Picasso, the best reform in corporate governance, why we might see a bifurcation of the Internet, what technology will explode in the the next 10 years, the most underrated media source, and more.”

Is Metoo Going Too Far by Waking Up with Sam Harris – “Sam Harris speaks with Rebecca Traister about her new book ‘Good and Mad: The Revolutionary Power of Women’s Anger’.”

Rational Feed

Scott:

Sort By Controversial by Scott Alexander – Spooky Fiction. [sorry no spoilers]

Rationalist:

Intellectual Status Isn’t That Different by Robin Hanson – Standard intellectual histories usually attribute breakthroughs to high status figures. However in most cases there were lower status people who took the positions first. There is unlikely to be much of an audience for honest intellectual histories because people care more about status games.

Long Legacies And Fights In An Uncaring Universe by Robin Hanson – Affecting the far future is very difficult in general. However Robin provides some strategies for reliably propagating changes forward in time.

AI:

Reinforcement Learning With Prediction Based Rewards by Open Ai – “We’ve developed Random Network Distillation (RND), a prediction-based method for encouraging reinforcement learning agents to explore their environments through curiosity, which for the first time1 exceeds average human performance on Montezuma’s Revenge. RND achieves state-of-the-art performance, periodically finds all 24 rooms and solves the first level without using demonstrations or having access to the underlying state of the game. RND incentivizes visiting unfamiliar states by measuring how hard it is to predict the output of a fixed random neural network on visited states. In unfamiliar states it’s hard to guess the output, and hence the reward is high. It can be applied to any reinforcement learning algorithm, is simple to implement and efficient to scale.”

Podcast:

Can We Avoid The Worst Of Climate Change by The Future of Life Institute – “Why the difference between 1.5 and 2 degrees C of global warming is so important, and why we can’t exceed 2 degrees C of warming. Why the economy needs to fundamentally change to save the planet. The inequality of climate change. Climate change’s relation to international security problems. How we can avoid the most dangerous impacts of climate change: runaway climate change and a ‘Hothouse Earth’. Drawdown’s 80 existing technologies and practices to solve climate change. ‘Trickle up’ climate solutions — why individual action is just as important as national and international action.”

Rational Feed

Rationalist:

The Great Stagnation by Bayesian Investor – “Cowen’s explanations for the stagnation are mostly unconvincing guesses that are somewhat less than half right. He seems to have entered the field, gathered the first few semi-ripe fruits he saw, and moved on without spotting the juicier fruit that was partly obscured by some leaves. Read Where Is My Flying Car? instead.”

AI:

Alignment Newsletter 30 by rohinmshah – Weekly AI newsletter. Features a long review of Paul Theory of aligning AI via iterated amplification. “The key insight is that humans are capable of decomposing even very difficult tasks into slightly simpler tasks. So, in theory, we could provide ground truth labels for an arbitrarily difficult task by a huge tree of humans, each decomposing their own subquestion and handing off new subquestions to other humans, until questions are easy enough that a human can directly answer them.”

Politics and Economics:

NBER Roundup by TrannyPornO – Weekly National Bureau of Economic Research roundup

NBER Roundup by TrannyPornO – Weekly National Bureau of Economic Research roundup

Podcast:

Michael Munger On Sharing Transaction Costs And Tomorrow 3.0 by EconTalk – “Munger analyzes the rise of companies like Uber and AirBnB as an example of how technology lowers transactions costs. Users and providers can find each other more easily through their smartphones, increasing opportunity. Munger expects these costs to fall elsewhere and predicts an expansion of the sharing economy to a wide array of items in our daily lives.”

Rational Feed

Rationalist:

How To Develop Scientific Intuition by Mark Eichenlaub – A discussion of how intuition develops. The autheor begins by discussing intuition in physics problem solving. He then moves on to a literature review of the more general case.

Podcast:

Peter Eckersley On Tough Choices On by Rationally Speaking – “Peter and Julia first delve into some of the most fundamental questions about privacy: What are the risks of losing privacy? Do we have more to fear from governments or industry? Which companies do a good job of protecting their users’ privacy? Are there tradeoffs between supporting privacy and supporting competitive markets? Next, they discuss Peter’s work measuring recent progress in AI, and debate to what extent recent progress is cause for optimism.”