Mega Links Post for Mar-Apr-May 2021

by ike17 min read31st May 2021No comments


World Modeling

It’s been a few months since my last links post, so here’s three months worth of catch-up. (formatting works better on substack)

Physics: “Physicist creates AI algorithm that may prove reality is a simulation” pretty sure this way oversells the paper (as with most of my physics links, see the Feb 2021 post) but still interesting. “Earth's Oxygen-rich Atmosphere will Last Only Another Billion years” “Meet Maxwell’s gambling demon—smart enough to quit while it’s ahead” “Factoring in gravitomagnetism could do away with dark matter” - explosive paper claims that we don’t need dark matter to explain galactic rotation curves. claims to identify a mistaken assumption in the dark matter paper. 

Sean Carroll @seanmcarroll

Dark matter exists -- an occasional reminder. The first evidence for dark matter came from the dynamics of galaxies and clusters of galaxies. But these days that is not our *best* evidence. 1/n

March 14th 2021

305 Retweets1,402 Likes

this tweet-thread goes over some of the evidence for dark matter. “Physics undergraduate proposes solution to quantum field theory problem” the Oumuamua mystery has allegedly been solved, and it’s not aliens (we think) “The LHCb Collaboration at CERN has found particles not behaving in the way they should according to the guiding theory of particle physics—the Standard Model.” Number go up: “Previous attempts to measure these minute differences between how atoms keep time — also referred to the ratio between them — had only ever delivered an accuracy of up to 17 digits.

But now using their new model, which includes the first-ever use of a ‘free-space link’ for this purpose (essentially, laser pulses of data going through the air instead of a cable,) the University of Colorado's BACON team (Boulder Atomic Clock Optical Network) has now measured this ratio reliably out to 18 digits.” 

Anders Sandberg @anderssandberg

#FridayPhysicsFun - the EmDrive has failed some fairly rigorous tests. So no neat reactionless space propulsion. But why were most people so confident that it really did not work even before this? 

Scientists Just Killed the EmDriveAfter failing critical thrust tests, the “impossible” engine has proven to be just

April 2nd 2021

4 Retweets43 Likes and

Toby Ord @tobyordoxford

The Edges of Our Universe I’ve just released a new paper exploring the largest-scale causal structure of our universe and its implications for what spacefaring civilisations could ever achieve.… 1/

April 6th 2021

131 Retweets645 Likes

and “The Edges of Our Universe” - survey of the various fundamental limits on how much of the universe we can theoretically access “Particle mystery deepens, as physicists confirm that the muon is more magnetic than predicted” “I show that quantum coherence is feasible over interstellar distances, and explain for the first time how astronomers can search for quantum transmissions sent by ETI to Earth, using commercially available telescopes and receiver equipment.” also links to which is good. A discussion of previous “discoveries” in physics that turned out to just be noise. Physics has so much data that there’s a ton of noise all the time. Various times it wasn’t aliens. “Hebrew University Researcher Introduces New Approach to Three-Body Problem, Predicts its Outcome Statistics.” “Physicists Prove That the Imaginary Part of Quantum Mechanics Really Exists” “How to Rewrite the Laws of Physics in the Language of Impossibility” “Time-Reversal Symmetry Breaking in a Superconductor” “How Gravity Is a Double Copy of Other Forces” “Quantum Entanglement Has Now Been Directly Observed at a Larger Macroscopic Scale” Superdeterminism makes sense of the quantum world by suggesting it is not as random as it seems, but critics say it undermines the whole premise of science. Does the idea deserve its terrible reputation? “Freshly Made Plutonium From Outer Space Found On Ocean Floor” “Unexpected 'Black Swan' defect discovered in soft matter for first time”

Philosophy: Cuttlefish can pass a version of the marshmallow test “Can Science Discover Moral Truths?” The Zen Anti-Interpretation of Quantum Mechanics - ‘Do “basketball games” really exist, or is that just a phrase we use to summarize our knowledge about certain large agglomerations of interacting quarks and leptons?’ “I Don’t Think We Have Any Free Will Whatsoever.” “Do Complex Numbers Exist?” “Quantum Mischief Rewrites the Laws of Cause and Effect” “Why Humans Fight” “Almost Everything You Do Causes Almost Everything” I think there’s some sense in which this is right, and some sense in which this is confusion over what we mean by causality. “What Should We Agree on about the Repugnant Conclusion?” “The hardest question ever asked: What is truth?” “On the evolutionary tree of plants, trees are regularly interspersed with things that are absolutely, 100% not trees. This means that, for instance, either:

The common ancestor of a maple and a mulberry tree was not a tree.

The common ancestor of a stinging nettle and a strawberry plant was a tree.

And this is true for most trees or non-trees that you can think of.”

Crypto: Claus Schnorr of Schnorr signature fame published a paper claiming to break RSA. Cryptographers are skeptical, see also “The Most Important Scarce Resource is Legitimacy” “People's Expensive NFTs Keep Vanishing. This Is Why” “While this change was hinted at in the original whitepaper published in 2017, the actual mechanism was not described. This was primarily due to the fact that we had no fucking clue how to implement it back then.”
The whole post is worth reading, especially the part about how they exploited someone trying to exploit the system: “Specifically what happened was the oracle detected a tx in the mempool trying to front run a price update. It then implemented a sandwich attack to raise the exchange fee to 99% for that transaction by sending one tx with higher gwei to raise it and another with lower gwei to drop it back down to the normal rate. Here is the transaction that slashed his funds by 99% and sent them to the feepool to be distributed back to SNX stakers.” “The Limits to Blockchain Scalability”  in depth analysis of the carry trade in crypto

AI: “Lying to the ghost in the machine” - current neural networks are pretty easy to deliberately trick “We started by simulating 72 fatal crashes as they occurred on public roads in our operating domain, which covers thousands of miles of road in southeast Phoenix.” “Pretrained Transformers as Universal Computation Engines” - I can’t make heads or tails of the abstract but it sounds like it’s doing something cool with pretraining “The Secret Auction That Set Off the Race for AI Supremacy” Open source attempt to replicate GPT-2 and 3. Can be used at “Tiny four-bit computers are now all you need to train AI” “Nine months since the launch of our first commercial product, the OpenAI API, more than 300 applications are now using GPT-3, and tens of thousands of developers around the globe are building on our platform. We currently generate an average of 4.5 billion words per day, and continue to scale production traffic.” “We fear and yearn for “the singularity.” But it will probably never come.” Semi-informative priors over AI timelines PyTorch + NumPy has a bug/feature related to random generation that hurts accuracy, and apparently quite a lot of projects are subject to it. “Shift to 7-nanometer process boosts the second-generation chip’s transistor count to a mind boggling 2.6-trillion” “This paper reports on the discovery of an accidental arbitrary code execution vulnerability in Marvin Minsky's 1967 implementation of the universal Turing machine.” “Advancing sports analytics through AI research” AI Safety research go brrr arguments that AI Risk is fake go brrr

Ethan Perez @EthanJPerez

Language models are amazing few-shot learners with the right prompt, but how do we choose the right prompt? It turns out that people use large held-out sets(!). How do models like GPT3 do in a true few-shot setting? Much worse: w/ @douwekiela @kchonyc 1/N

May 24th 2021

99 Retweets449 Likes

I don’t quite understand this paper, but it seems to be arguing that GPT-3 and related work looks better than it should on various metrics because of something to do with the way parameters are tuned. 

General: “Court exam adjourned after witness, defendant found together in livestream” “He Got $300,000 From Credit-Card Rewards. The IRS Said It Was Taxable Income.” “It is not yet clear if Belgian authorities plan to claim the reward.” Some reverse engineered data on NYT A/B testing “A Bird-Feed Seller Beat a Chess Master Online. Then It Got Ugly” (they were accused of cheating and banned) “Can we stop hurricanes?” “Trader Buys $36 Million of Copper and Gets Painted Rocks Instead” “Here's something that I found really fascinating: There was not a ton of what I would consider rigorous critical work about the meaning of blonde.” Ray Dalio says bonds are a bad investment (as of mid-Mar 2021) How to hack the California tax system: “You may donate to that SBE member who will vote against you. This may sound counterintuitive, but the idea is that both you and the SBE member must then disclose that contribution. Any contribution of $250 or more must be disclosed. Your contribution will disqualify that SBE member from considering your case. The only exception is if the SBE member returns the contribution within 30 days from the time he or she knows, or has reason to know, of the contribution. Often, though, a contribution will not be returned.” That time they tricked reality show contestants into thinking they went to space some testing showing that Apple’s new M1 chip is pretty efficient at multiplying 128 bit integers “America’s Covid Swab Supply Depends on Two Cousins Who Hate Each Other” anti-piracy company asks for exemption from the DMCA so they can track down pirates better “Skylab: The myth of the mutiny in space” “$500,000 Jefferson Davis chair stolen in Selma will be a toilet unless Confederate group hangs banner, email claims” “A software mistake caused a Tui flight to take off heavier than expected as female passengers using the title “Miss” were classified as children, an investigation has found.” “This caused the load sheet – produced for the captain to calculate what inputs are needed for take-off – to state that the Boeing 737 was more than 1,200kg lighter than it actually was.” Why Historians Are Reexamining the Case of the Woman Who Gave Birth to Rabbits Apparently NBA players get challenged all the time by people that are way overconfident about their abilities. “They Hacked McDonald’s Ice Cream Machines—and Started a Cold War” really wild story about a con man who “bought” a hotel and screwed over the sellers. “And yet, to not speak out was wrong. To destroy a billion dollars is to destroy an almost unimaginable amount of human well-being. Spent carefully on anti-malarial bed nets and medicine, one billion dollars could save a million lives. This was a crime, and failing to try and stop it would be as bad as committing it myself. And if I could not prevent it, then what reason was I being paid such a high salary? How could I justify my income if not by prevailing in situations such as these?” Snowden got invited to a real estate virtual conference, did some digging and realized the host was a scammer, so he went on and called him out live. Card grading services are super backlogged due to a surge in demand. turns out if you succeed bigly enough, you reach the highest number and win capitalism. Good for Warren Buffet! someone was able to take over a website by using a fake court order to get the registrar to hand it over. 

Archegos, a 100B fund that nobody heard about, blew up and caused big losses for a bunch of wall street banks. The first story I saw was about the banks selling off massive block trades and nobody knew what was going on:, followed by new info trickling out. different banks got out at different times, some broke even, some lost billions. talks about how Bill Hwang went from under a billion to a 10B+ fortune in a very short period of time, followed by a spectacular collapse. “He has been quoted in hundreds of articles and television broadcasts as a member of the public (that is, a "man on the street" rather than a newsmaker or expert)” “A Somewhat Comprehensive History Of U.S. Senators Who Have Died In Duels” interesting discussion of how investment goes wrong, in between pitching his startup Tldr: sign up to donate bone marrow, then send a GDPR request for your DNA data "I have unlimited money to blow on Teslas. If you take away my Tesla, I will get another Tesla. That's how it works," Sharma said. The most bizarre dispute of recent years involved a debate about who was to blame for a train hitting a peacock. If it was defined as a small bird, then the company driving the train was responsible: if it was categorised as a large bird, then the blame went to the operator of the tracks. The two sides ended up haggling over whether peacocks were bigger than geese. (The answer: a peacock is a “large bird”.)

Math: “New Algorithm Breaks Speed Limit for Solving Linear Equations” “Factoring 2048 RSA integers in 177 days with 13436 qubits and a multimode memory” “A paper posted in October comes the closest yet, describing the fastest-ever method for multiplying two matrices together. The result, by Josh Alman, a postdoctoral researcher at Harvard University, and Virginia Vassilevska Williams of the Massachusetts Institute of Technology, shaves about one-hundred-thousandth off the exponent of the previous best mark. It’s typical of the recent painstaking gains in the field.” “In this paper, I present and motivate a modal set theory consistent with the idea that there is only one size of infinity.”


Lise Eliot @Lise_Eliot

1/🧵 Tweetorial on our new paper, “Dump the Dimorphism: Comprehensive synthesis of brain studies finds few male-female differences beyond size.” (…). It's big (43K words, 616 refs) so I'll cut to the chase.

Dump the “dimorphism”: Comprehensive synthesis of human brain studies reveals few male-female differences beyond sizeWith the explosion of neuroimaging, differences between male and female brains have been exhaustively analyzed. Here we synthesize three decades of hu…

March 20th 2021

537 Retweets1,174 Likes

tweet-thread about a paper finding few differences between male and female brains “Did the Black Death Rampage Across the World a Century Earlier Than Previously Thought?”

Liam Bright @lastpositivist

New paper by @haicinnamon and myself out. As per our title, in it we claim that Scientific Conclusions Need Not Be Accurate, Justified, or Believed by their Authors. Of course *our* conclusion is accurate, justified, and believed by the authors.

April 15th 2021

27 Retweets135 Likes

“We argue that the main results of scientific papers may appropriately be published even if they are false, unjustified, and not believed to be true or justified by their author.”, via

Sarah Constantin @s_r_constantin… whoa this is fascinating.

Is This the Most Interesting Idea in All of Science?Neurons might contain something incredible within

April 16th 2021

40 Retweets235 Likes “How Long Can We Live?

New research is intensifying the debate — with profound implications for the future of the planet.” “If I fits I sits: A citizen science investigation into illusory contour susceptibility in domestic cats (Felis silvestris catus)” People who live past 105 years old have genes that stop DNA damage “At RAND in 1954, Armen A. Alchian conducted the world's first event study to infer the fuel material used in the manufacturing of the newly-developed hydrogen bomb. Successfully identifying lithium as the fusion fuel using only publicly available financial data, the paper was seen as a threat to national security and was immediately confiscated and destroyed.“ “Perverse Downstream Consequences of Debunking: Being Corrected by Another User for Posting False Political News Increases Subsequent Sharing of Low Quality, Partisan, and Toxic Content in a Twitter Field Experiment” “Plants Produce Ultrasonic Clicks under Stress”

Econ/Psychology: , thread at

Consequentialist Libertarian @LibertarianWonk

NEW Piece from me, why wage 'stagnation' is a myth: 

The Wage Stagnation MythIf you have been remotely in touch with modern day politics, then you have probably heard of the claim that wages in the US (and other developed nations) have been ‘stagnating’ for the past few…

March 23rd 2021

19 Retweets96 Likes

, arguing that wage stagnation is a myth

scott cunningham @causalinf

This new article in economic inquiry coauthored by @nickchk is absolutely terrifying. Seven competent people were given two causal studies to replicate from scratch. No two people even report the same sample size and the standard deviation across estimates was 4x SE of original!

March 24th 2021

395 Retweets1,304 Likes

“Seven competent people were given two causal studies to replicate from scratch. No two people even report the same sample size and the standard deviation across estimates was 4x SE of original!” replication crises go brrr. Original paper at Critique of Elizabeth Warren’s book The Two Income Trap “Which brings me to the reason experts should be more reluctant to lie to the public: They aren’t experts on the topic of when to lie.” 

Lyman Stone 石來民 🦬🦬🦬 @lymanstoneky

Q: What is the wrongest thing you could ever do in economics A: Publish a paper saying shipping supply shocks are unimportant drivers of shipping costs the week after The Boat Got Stuck #NBERday 

Dry Bulk Shipping and the Evolution of Maritime Transport Costs, 1850-2020Founded in 1920, the NBER is a private, non-profit, non-partisan organization dedicated to conducting economic research and to disseminating research findings among academics, public policy makers, and business

March 30th 2021

7 Retweets44 Likes Noah Smith argues against Jason Hickel’s worldview. Max Roser also has similar issues with Hickel. I found it pretty interesting that one of Roser’s points is that $10/day is a overly low number for poverty (see

Max Roser @MaxCRoser

@Noahpinion Had a long discussion with him this morning about it. A wild discussion. But we did at least finally understand why we disagree about the need for growth: He believes that adjusted for prices in the country, $10 per day is a "very high poverty line".

Max Roser @MaxCRoser

@jasonhickel @JKSteinberger Ah thank you very much Jason. That explains it. Then we disagree about our notion of who a poor person is. I do believe a person who lives on just $20 or $25 per day is a poor person. Your poverty line of $10 is a third of the price of your own book. That’s very little.

April 2nd 2021

4 Retweets44 Likes

), but one of Noah’s points is that increases in the level of low-income people matter more than Hickel admits. There’s some tension there. “Estimating the college wealth premium: Not so easy” “Information Markets Create A New Asset Class” (see also my previous post,

Cailin O'Connor @cailinmeister

New Paper! We find that subjects think you are less at risk of COVID infection when engaged in morally good actions, and more likely to catch COVID while doing morally bad things. In other words, risk judgments are systematically skewed.

May 7th 2021

581 Retweets1,982 Likes

“We find that subjects think you are less at risk of COVID infection when engaged in morally good actions, and more likely to catch COVID while doing morally bad things.  In other words, risk judgments are systematically skewed.” Turns out that Texas reopening didn’t really matter for either the economy or for covid. Nvidia continues their war on miners using retail GPUs, cutting the hashrate for more models so they can’t be used for mining efficiently. Kahneman and co’s new book Noise has some embarrassing mistakes. many things that look like risk aversion can be explained better by other theories. 

Lab leak:

I’ve collected various articles that looked interesting on the topic of whether covid leaked from a lab. I’m still not sure what to think, and haven’t had the time to do a deep dive, but I figured I’d put all the articles I had saved in one place. argues that Josh Rogin misrepresented US cables and took them out of context, thread at

Zichen Wang @ZichenWanghere

~ @joshrogin last year exclusively reported U.S. "cables warned of safety issues at Wuhan lab studying bat coronaviruses" It "launched claims that coronavirus escaped from Chinese lab," according to @washingtonpost Here's my takedown by @ZichenWanghere 

Josh Rogin’s WashPo column & book misrepresented a U.S. diplomatic cable from WuhanSetting the record straight with the Washington Post Opinion

April 2nd 2021

34 Retweets131 Likes comments on

Alina Chan @Ayjchan

I’m putting together a primer pack of must-read layperson articles for anyone who’s just getting into the topic of whether SARS-CoV-2 could have come from a lab in Wuhan. In no order of preference or chronology: @rowanjacobsen @BostonMagazine 

Could COVID-19 Have Escaped from a Lab?The Broad Institute’s Alina Chan believes it could

May 20th 2021

110 Retweets259 Likes

thread with a bunch of articles 

Pwn All The Things @pwnallthethings

Might be time for the semi-regular "articles that make claims based on paraphrases of supposedly leaked US intelligence are nearly worthless" thread again

May 23rd 2021

1,614 Retweets4,163 Likes

thread responding to the WSJ “3 Wuhan scientists were sick” article

Nsikan Akpan, PhD @MoNscience

The Intergovernmental Panel on Climate Change first met in 1990, where it predicted that GHGs could fuel global warming. Even as more evidence accumulated, efforts to sow doubt delayed action. The “natural origins” vs “lab leak” debate around COVID-19 has entered this realm...

May 22nd 2021

441 Retweets1,054 Likes

thread arguing against lab leak hypothesis 


New Comment