Not Quite a Blog

On 5 January, we will have two runoff elections for control of the Senate. I want Democrats to win (and have donated to them), but when it comes to forecasting and profit-making, my only interest is accurate views.

I worked hard over the past few weeks to help start Open Model Project, which produced a transparent model of the Georgia Senate runoff election.

First, two meta-predictions about our model:

• If you start with default assumptions and deviate only by enabling the COVID adjustment, you will get a more accurate result - 70%
• If you start with default assumptions and deviate only by removing Trafalgar, you will get a more accurate result - 60%
  

The remaining predictions will come from the model as modified by the two above changes.

Now some actual predictions about the races:

• Republicans hold the Senate (51+ seats) - I personally give this 61% odds and on betting markets I am risking $1000 to earn $887 after fees. This betting gives an implied odds of 53% and an expected value of $151.07.

• Perdue wins - 59% odds, risking $1144.43 to earn $849.86 (57%, EV $32.20)

• Loeffler wins - 52% odds, risking $594.69 to earn $537.18 (53%, EV -$6.12). Yes this expected value is negative because I made a lot of bets when my personal odds for Warnock were lower. I am holding these bets because the markets have surged for Warnock and it is better EV to hold than to sell.

• No split (Ds or Rs win both races) - 87% odds, risking $1000 to earn $288 (78%, EV $120.56)

• Warnock wins by 0-2% or Loeffler wins by 0-3% - 58% odds, risking $492.70 to earn $371.16 (57%, EV $8.33)
  

• Ossoff wins by 0-2% or Perdue wins by 0-3% - 58% odds, risking $695.18 to earn $269.65 (72%, EV -$135.58). See above about negative EV. I also own this position to help hedge risk.
  

• Loeffler does not have fewest votes among all four candidates - 76% odds, risking $295.71 to earn $146.96 (67%, EV $40.72)
  

• Loeffler has the most votes among all candidates - 24% odds, risking $94.73 to earn $535.27 (15%, EV $56.47)

Also here is where I am at for turnout:

• Turnout is less than 3.6M - 1% odds, will lose $1985.74 if this occurs
• Turnout is 3.6-3.8M - 6% odds, lose $197.74
  
• Turnout is 3.8-4M - 40% odds, earn $698.55
• Turnout is 4-4.2M - 47% odds, earn $466.77
• Turnout is 4.2-4.4M - 4% odds, lose $197.74
• Turnout is more than 4.4M - 1% odds, lose $1772.08

(EV = $434.42)

Some scenario analysis:

• Expected value based on my given odds: +$820.62
• Expected value if Warnock + Ossoff both win by >2% (turnout 3.6-3.8M): -$5515.18
• Expected value if Warnock + Ossoff both win by 0-2% (turnout 3.8-4.0M): -$1502.20
  
• Expected value if Warnock + Perdue win (turnout 3.8-4.0M): +$1091.09
  
• Expected value if Loeffler + Ossoff win (turnout 3.8-4.0M): +$1301.34
  
• Expected value if Loeffler + Perdue win (turnout 4.0-4.2M): +$3731.85

Some caveats:

• I already have lost $427.11 on net in these above bets, changing some of my positions over time. This is not factored into the scenario analysis above.
• While this is after fees, this does not include the PredictIt withdrawal fee of 5%.
• My odds given here are odds that I give as of today, not necessarily the odds as of the time I made the bet.
• Loss and expected value assume I don’t sell out early, which I might to mitigate loss.
• Nothing in this post should be construed as investment advice.

On 6 January, Congress will meet to certify the electoral college. This is usually a formality, but it looks like this time there will be some drama. Because PredictIt and Polymarket let us take bets on institutional failure and because I want to earn money and improve my forecasting skills, here are my thoughts.

To start, Biden will be inaugurated as President. I put >99% odds on this and I have risked $8068 on the betting markets to potentially earn $855 after fees, for implied odds of 90% and an expected value of $855 after fees.

There will be an objection to the electoral college votes - I put 97% odds on this and I have risked $849.30 to earn $40.23 after fees (95% implied odds, expected value: $13.54 after fees)

How many states will be subject to challenge?

• No states are challenged - 3% odds, will lose $522.99 if this occurs
• One state - 5% odds, lose $297.83
• Two states - 10% odds, lose $350.94
• Three states - 33% odds, earn $1453.81
• Four states - 21% odds, earn $544.81
• Five states - 4% odds, lose $380.79
• Six states - 20% odds, lose $139.59
• 7+ states - 4% odds, lose $401.48

(EV = $469.28)

How many Senators will sign objections?

• Objection is signed by no senators - 3% odds, will lose $635.57 if this occurs
• One senator - 9% odds, will earn $242.85 if this occurs
• Two senators - 8% odds, earn $242.85
• Three senators - 25% odds, earn $1336.05
• Four senators - 17% odds, earn $221.82
• Five senators - 3% odds, lose $634.99
• Six senators - 20% odds, earn $1149.81
• 7+ senators - 15% odds, lose $406.12
  

(EV = $543.93)

Who will sign? I think Mike Lee does not sign the objection - 95% odds, risking $849.14 to earn $59.27 (93%, EV $13.85). The others I am unsure about.

How many representatives will sign objections?

• Objection is signed by no representatives - 3% odds, will lose $0.63 if this occurs
• 1-3 reps - 43% odds, will earn $3326.21 if this occurs
• 4-6 reps - 30% odds, earn $2785.45
• 7-9 reps - 2% odds, earn $865.90
• 10-24 reps - 2% odds, earn $20.74
• 25+ reps - 20% odds, lose $656.35

(EV = $2152.35)

What about the actual vote?

I think that more than 100 and less than 160 Representatives vote to reject the results - 90% odds, risk $694.32 to earn $232.18 (75%, EV $139.53)

also more than 10 Senators vote to reject the results - 90% odds, risk $681.83 to earn $286.93 (70%, EV $190.05)

Lastly, I think Pence will preside over the certification - 90% odds, risk $452 to earn $83.70 (84%, EV $30.13)

Some scenario analysis:

• Expected value if I lose every bet except the >99% ones (amount at stake): -$3454.15
• Expected value if I am rather unlucky/wrong: $504.14
• Expected value if I am somewhat unlucky/wrong: $1482.95
  
• Expected value of all my bets if I am perfectly right about the odds: $4407.66
  
• Expected value if I win every bet: $6537.59
  

Some caveats:

• I have already lost $1249.43 on net in these above bets, changing some of my positions over time. This is not factored into the scenario analysis above.
• While this is after fees, this does not include the PredictIt withdrawal fee of 5%.
• My odds given here are odds that I give as of today, not necessarily the odds as of the time I made the bet.
• Loss and expected value assume I don’t sell out early, which I might to mitigate loss.
  
• Nothing in this post should be construed as investment advice.

• Transfer files over from backup disk
• Get Chrome
• Set up Chrome to cloud sync (this auto-installs extensions)
• Get iTerm 2
• Clean up Dock
• Get iTunes working
• Max keyboard key repeat and minimize delay until repeat
• remap caps lock to ESC
  
• Max trackpad mouse speed
• Disable “swipe between pages” gesture
• Auto-hide dock and add magnification
  
• Show battery percentage
• Install Alfred
• Install Rectangle
• Install homebrew
• brew install wget
• setup https://github.com/peterhurford/dotfiles
• Copy ssh keys and ~/.secret_zshrc
• git config –global –edit
• Set up Slack + sign into workspaces
• Get Flux
  
• Install copyclip
• Install Skype
• Install Gimp
• Install VLC
• Disable all smart features from TextEdit
  
• brew install Python
• brew install pyenv
  
• pip install virtualenv
• pip install virtualenvwrapper
• rebash
• bootpython
• mkvirtualenv dev
• brew Install R
  
• Install Heroku toolbelt
• Install Jupyter notebook -> PDF
• Sign into all emails
  

Some of these are classics. Some of these I made up. Some of these are insightful. Some of these don’t make sense. Enjoy!

Acton’s Dictum - “Power tends to corrupt, and absolute power corrupts absolutely.”

Alpha Eta’s Motto - “The best for us”

Altman’s Maxim - “I always want […] a project that, if successful, will make the rest of my career look like a footnote.”

Amara’s law - “We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.”

Betteridge’s law of headlines - “Any headline that ends in a question mark can be answered by the word no.”

Bogosian’s Law - “For every hypocrisy, there is an equal and opposite hypocrisy”

Brook’s Law - “adding manpower to a late software project makes it later”

Bryce’s Maxim - “There is nothing more unproductive than to build something efficiently that should not have been built at all.”

Campbell’s Law - “The more any quantitative social indicator is used for social decision-making, the more subject it will be to corruption pressures and the more apt it will be to distort and corrupt the social processes it is intended to monitor”

Chekov’s gun - If you see a gun in a TV show, it is going to be used.

Chesterton’s fence - Avoid changing things until you understand why they are the way they are.

Clarke’s First Law - “When a distinguished but elderly scientist states that something is possible, he is almost certainly right. When he states that something is impossible, he is very probably wrong.”

Clarke’s Second Law - “The only way of discovering the limits of the possible is to venture a little way past them into the impossible.”

Clarke’s Third Law - “Any sufficiently advanced technology is indistinguishable from magic.”

Cromwell’s Law - “0% and 100% are not valid probabilities”

Cunningham’s Law - “The best way to get the right answer on the Internet is not to ask a question, it’s to post the wrong answer.”

The Date-Time Continuum - Don’t make plans with a date more than X days in the future, where X is the number of days you have been dating.

Drucker’s Maxim - “What gets measured, gets managed”

Emmy the Great’s Maxim - “You’re not unlucky, you’re just not very smart”

The Front Porch Test - Evaluate a decision based on how you will reflect upon it when you are an 80-year-old resting on your front porch.

Gall’s Law - “A complex system that works is invariably found to have evolved from a simple system that worked.”

Gell-Mann amnesia - Believing articles outside one’s area of expertise, even after acknowledging that neighboring articles in one’s area of expertise are completely wrong

Gibson’s Law - “For every PhD there is an equal and opposite PhD.” (There are experts on both sides of any issue.)

Godwin’s Law - “as an online discussion grows longer, the probability of a comparison involving Nazis or Hitler approaches 1.”

Goodhart’s Law - “When a measure becomes a target, it ceases to be a good measure.”

Hanlon’s Razor - “Never attribute to malice that which can be adequately explained by stupidity.”

Hitchen’s Razor - “What can be asserted without evidence can also be dismissed without evidence.”

Hofstader’s Law (planning fallacy) - “It always takes longer than you expect, even when you take into account Hofstadter’s law”

The Hot-Crazy Line - You’re willing to tolerate more craziness from a more physically attractive person.

Jinn’s Maxim - “there’s always a bigger fish”

Joy’s Law - “No matter who you are, most of the smartest people work for someone else”

Kerckhoffs’s Principle / Shannon’s Maxim - Assume your enemy knows everything about your security system (i.e., don’t practice “security through obscurity”)

Le Mis’s Maxim - “At the end of the day you’re another day older”

Linch’s Law - Any forecaster who tries to forecast his or her own number of Twitter followers, will eventually encounter an event that inflates their follower count.

Litany of Gendlin - “What is true is already so. Owning up to it doesn’t make it worse. Not being open about it doesn’t make it go away. […] People can stand what is true, for they are already enduring it.”

Litany of Hodgell - “That which can be destroyed by the truth should be.”

Litany of Jai - “Almost no one is evil. Almost everything is broken.”

Litany of Lenin - “There are decades where nothing happens; and there are weeks where decades happen”

Litany of Tarski - “If the box contains a diamond, I desire to believe that the box contains a diamond; If the box does not contain a diamond, I desire to believe that the box does not contain a diamond; Let me not become attached to beliefs I may not want.”

Littlewood’s Law - A person can expect to observe a miracle about once a month.

Marcus’s Law - As time spent running a business goes longer, the probability of needing to become an expert on currency exchange rates approaches 1

Metcalfe’s Law - The value of a network is proportional to the square of the number of members of the network.

Moore’s Law - “the complexity of integrated circuits doubles every 24 months”

Mosteller’s Maxim - “It is easy to lie with statistics, but easier to lie without them”

Murphy’s Law - “Anything that can go wrong will go wrong.”

Nate’s Law - The polling error is always in the opposite direction that the conventional wisdom expects.

Newton’s Flaming Laser Sword - “What cannot be settled by experiment is not worth debating.”

Occam’s Razor - “When two or more explanations are offered for a phenomenon, the simplest full explanation is preferable.”

O’Sullivan’s First Law - “All organizations that are not actually right-wing will over time become left-wing.”

Pareto Principle - “80% of consequences stem from 20% of the causes”

Parkinson’s Law - "work expands so as to fill the time available for its completion”

Asimov’s corollary - “In ten hours a day you have time to fall twice as far behind your commitments as in five hours a day.”

Parkinson’s Law of Triviality - “The time spent on any agenda item will be in inverse proportion to the sum of money involved.”

Poe’s Law - “it is utterly impossible to parody […] in such a way that someone won’t mistake for the genuine article”

Postel’s Law - “Be conservative in what you do, be liberal in what you accept from others”

Pournelle’s Iron Law of Bureaucracy - “In any bureaucratic organization there will be two kinds of people: those who work to further the actual goals of the organization, and those who work for the organization itself. Examples in education would be teachers who work and sacrifice to teach children, vs. union representatives who work to protect any teacher including the most incompetent. The Iron law states that in all cases, the second type of person will always gain control of the organization, and will always write the rules under which the organization functions.”

Putt’s Law - “Technology is dominated by two types of people, those who understand what they do not manage and those who manage what they do not understand.”

Rossi’s Iron Law - “The expected value of any net impact assessment of any large scale social program is zero”

Rossi’s Stainless Steel Law - “the better designed the impact assessment of a social program, the more likely is the resulting estimate of net impact to be zero.”

Sagan’s Standard - “extraordinary claims require extraordinary evidence”

Say’s Law - supply creates its own demand

Schneier’s Law - “Any person can invent a security system so clever that she or he can’t think of how to break it.”

Shannon Principle - Assume success will occur at the 50% mark

Shooster’s Law - “Anything meat companies don’t want to happen is likely good for animals”

Sinatra’s Maxim - “The best is yet to come.”

Streisand Effect - Attempts to hide, remove, or censor information often have the unintended consequence of further publicizing that information

Sturgeon’s Law - “Ninety percent of everything is crap”

Sutton’s Law - “One should first consider the obvious” / “When you hear hoofbeats, think horses, not zebras.”

Swift’s Law - “If you play stupid games, you will win stupid prizes.”

Wittgenstein’s Ruler - “When you measure a table with a ruler, you are also measuring the ruler with the table.” (That is, you must take into account error in your measurement system.)

The Iowa Caucus for the Democratic Primary is today. What’s the best forecast? Do prediction markets add in additional information to make more accurate forecasts or just introduce noise? Are models better than markets? While we can’t really generalize from one example*, I decided to collect a bunch of forecasts as of 5pm ET prior to the Iowa Democratic Caucus, stated as the odds of a particular candidate winning in Iowa. We can later see who predicts the best.

538

Sanders - 41%
Biden - 32%
Buttigieg - 14%
Warren - 10%
Klobuchar - 2%
All others - 1%

NYT Upshot

Sanders - 35%
Biden - 28%
Buttigieg - 18%
Warren - 13%
Klobuchar - 6%
All others  - <1%

Harry Enten

Sanders - 37.5%
Biden - 25%
Buttigieg - 17.5%
Warren - 15%
Klobuchar - 4%
All others - 1%

Real Clear Politics Odds Average**

Sanders - 34.3%
Biden - 28.5%
Bloomberg - 13.7%
Warren - 6.6%
Buttigieg - 4.6%
All Others - 12.3%

Good Judgement Open

Sanders - 41%
Biden - 37%
Buttigieg - 11%
Warren - 10%
All others - 1%

Election Betting Odds

Sanders - 60.6%
Biden - 18.1%
Buttigieg - 10.2%
Warren - 6.6%
Klobuchar - 3.1%
All others - 1.4%

PredictIt***

Sanders - 56%
Biden - 17%
Buttigieg - 9%
Warren - 6%
Klobuchar - 3%
All others - 9%

BetFair****

Sanders - 63%
Biden - 19.6%
Buttigieg - 6.8%
Warren - 5.6%
Klobuchar - 0.2%
All others - 4.8%

Endnotes

*Ideally we would try to gather similar data from many other events to form a track record across a lot  more than one datapoint, but that will be an exercise for another time.

**Sums to 115.8% by default, so I normalized predictions to sum to 100%.

***Sums to 124% by default, so I normalized predictions to sum to 100%.

****Average of back + lay

Welcome to Peter’s “Links When I Feel Like It”, where periodically (as I feel like it), I take my link trawler through the vast oceans of the internet and curate for you only the finest link fish for your consideration.

Please note that I do not necessarily agree or fully endorse all links contained here - I just think they’re worth thinking about.

Without further ado, throughout all of time and space, here are the links:

PANDEMIC!!! [GimletMedia]: “If a pandemic ripped across the world, how bad would it really get? You’ve heard the horror stories, but you’ve never heard one like this. Dr. Anthony Fauci, who advises the President on emerging infectious diseases, helps us out.”

Why we panic about coronavirus, but not the flu [Axios]: “If you’re freaking out about coronavirus but you didn’t get a flu shot, you’ve got it backwards. The big picture: A novel outbreak will always command more attention than a common illness, and the coronavirus is a serious health threat. But our newfound hyper-vigilance about infections might be more helpful if we could redirect some of it toward influenza — a significantly deadlier virus that strikes every year.”

It’s time for the doom-mongers to clock off [Unherd]: “It is, though, absolutely ridiculous that the Doomsday Clock considers the very real problems of modern society — Trump being mad, Putin being machiavellian, the climate getting dangerously warm — as in any way as great an imminent and deadly threat as, say, the moment in 1952 when the USA and USSR tested their first thermonuclear devices within a few months of each other. And, yet, the Bulletin put that at just two minutes to midnight. This is not scientific risk assessment: this is PR.”

Doing good is as good as it ever was [EAForum]: “Try to have realistic expectations about how much good you can do and get satisfaction from that. There are lots of important problems left. There are as big and as important as they ever were, and the world needs your contributions just as much as before.”

The EA events ecosystem: How to get more involved (as a participant) [EAForum]: Want to get more involved in the effective altruism community? Find an EA meetup on the Effective Altruism Hub, sign up for the EA Pen Pals, look at profiles on the EA Hub, use the EA Forum, look to WANBAM (Women and Non-Binary Altruism Mentorship) if relevant, go to an EAGx conference, and/or apply to EA Global.

If you want to follow your dreams, you have to say no to all the alternatives [Oliveremberton]: “People never want to do one thing. We want to do all the things. We simultaneously want to exercise and to learn Spanish and to go out for pizza. Our desires are countless, independent agents, working to nudge our beachball in their own selfish direction. And so usually, that ball is going nowhere. It’s controlled more by the terrain than by the will of what’s inside it. […] Let’s fix that.”

What Makes The Top 10% Of Founders Different? [YouTube]: From observing ~2500 founders at YC, top 10% founders are consistently able to declare what they want to do within a two-week interval, do all of it, and learn something from it. They may not always do the right thing or do it correctly, but they don’t get stuck in the execution step. Top 10% founders are also really good at communication to investors, employees, customers, etc. - they’re able to explain their business to anyone at any level of understanding, without stumbling. Lastly, top 10% founders can persevere in the face of defeat and failure. Quality of ideas are hard to evaluate in the beginning and should be used sparingly to evaluate founders.

Who Should Founders Listen To For Advice? [YouTube]: There’s an overwhelming amount of advice out there, so you need to also pay attention to who you do and don’t listen to for advice. Be careful of asking for advice too much, as it takes away time from actually doing things. Be careful of trying to just seek advice until you get the answer you want, as this is a waste of time. Be careful of trying to average a bunch of advice, because this could steer you away from the outlier advice that would make you truly great. Try to pick just a few trusted advisors, avoiding people who are overly prescriptive, people who only have experience with one particular company, or jerks. Remember that there are a lot of different ways to do things that will work.

Coping with Founder Depression [YouTube]: Depression among founders is normal. Nothing ever goes right in a company all the time. What is important is to acknowledge it and be proactive. Talk to friends and be honest about things. It may be especially helpful to find trusted friends outside the start-up scene. Also, if you think other people may be struggling with things you have gone through, try to be there for them.

What Bloomberg’s $11 million Super Bowl ad would cost you on your budget [WashingtonPost]: “Mike Bloomberg spent $11 million of his own money on a 60-second Super Bowl ad. That was 0.018 percent of his net worth. For you, that’s equivalent to $22, a 14-inch Domino’s Buffalo Chicken pizza.”

She Helped a Customer in Need. Then U.S. Bank Fired Her. [NYTimes]: “When young Americans say in polls that they react more positively to ‘socialism’ than to 'capitalism,’ it’s because of the hypocrisy of institutions like U.S. Bank. I’ve often noted that companies have enormous capacity to help their communities. But too often they act like American tobacco companies, which killed more people than Stalin did, or pharma companies peddling opioids, or McKinsey & Company advising a business to 'get more patients on higher doses of opioids,’ or Boeing mocking regulators.”

Donald Trump, We Are Coming For You [CurrentAffairs]: “The problem is not so much that other Democrats “focus” on Trump himself rather than on an agenda for working people. It is that they are incapable of making the anti-Trump case that needs to be made. He’s not a clown. A clown is funny, entertaining. He’s a monstrosity. None of this is entertaining. It’s just horrific. People are dead. So what we have to have is a politics that says: There are now going to be consequences for waging war on working people.”

What Does It Mean To Be A Socialist? [CurrentAffairs]: “We have to build a movement, we have to all get along, but we have to do it without compromising our principles. Study those who came before us. Look at what they did, understand ourselves as continuing their work. Think strategically. Avoid cargo cult politics. This is an experiment. We have to undertake the long, slow process of figuring out how to organize successfully. Maintain clarity of vision. Constantly remind ourselves exactly what it is we are fighting for. Make sure the right questions are being asked. Seize the moment. Believe in ourselves.”

The radical moral implications of luck in human life [Vox]: “Neither human genes nor human societies distribute life’s gifts according to any principle we would recognize as fair or humane, given the extraordinary role of luck in our lives. We all become adults with wildly different inheritances, starting our lives in radically different places, propelled toward dramatically different destinations. We cannot eliminate luck, nor achieve total equality, but it is easily within our grasp to soften luck’s harsher effects, to ensure that no one falls too far, that everyone has access to a life of dignity. Before that can happen, though, we must look luck square in the face.”

Joe Biden is the only candidate with a real shot at getting things done [Vox]: Biden has the best shot at carrying the Senate to actually implement his ideas. While he may not pass progressive purity tests, he’d still be the most progressive Democratic nominee in history if he won. Also, Trump has been running — and losing — against Biden for months.

Language: A Key Mechanism of Control [Fair]: “Often we search hard for words to help us define our opponents. Sometimes we are hesitant to use contrast. Remember that creating a difference helps you. These are powerful words that can create a clear and easily understood contrast. Apply these to the opponent, their record, proposals and their party. […] Use the list below to help define your campaign and your vision of public service. These words can help give extra power to your message. In addition, these words help develop the positive side of the contrast you should create with your opponent, giving your community something to vote for!”

The Facebook Trials: It’s Not “Our” Data [MarginalRevolution]: “Facebook hasn’t taken our data—they have created it. Facebook and Google have made billions in profits, but it’s utterly false to think that we, the users, have not been compensated. Have you checked the price of a Facebook post or a Google search recently? More than 2 billion people use Facebook every month, none are charged. Google performs more than 3.5 billion searches every day, all for free. The total surplus created by Facebook and Google far exceeds their profits.”

“Theory vs. Data” in statistics too [Noahpinion]: “[W]hen you have solid, reliable structural theory or you need to make predictions about situations far away from the available data, use more theory. When you don’t have reliable theory and you’re considering only a small change from known situations, use less theory. This seems like a general principle that can be applied in any scientific field, at any level of analysis (though it requires plenty of judgment to put into practice, obviously).”

What Really Went Wrong in Afghanistan [TheDispatch]: “In December, the Washington Post published “Afghanistan Papers,” a series of investigative reports on the war[. …] As a former director for Afghanistan and Pakistan on the National Security Council staff under Presidents George W. Bush and Barack Obama, I was among those interviewed by SIGAR. […] I strongly disagree with much of the Post’s analysis. Some of the Post’s investigative reports lack context, perspective, or depth. Some of the Post’s claims are simply false. The Post misdiagnoses what went wrong in the war, while the actual main mistakes of the war—Bush’s light footprint and Obama’s withdrawal deadline—get comparatively little mention.”

Into the Personal-Website-Verse [Matthiasott]: “There is one alternative to social media sites and publishing platforms that has been around since the early, innocent days of the web. It is an alternative that provides immense freedom and control: The personal website. It’s a place to write, create, and share whatever you like, without the need to ask for anyone’s permission. […] But maybe the most compelling reason why a personal website and also learning how to build one is incredibly valuable is this: community.”

A Theory of Law [Greenbag]: “It is a common practice among law review editors to demand that authors support every claim with a citation. These demands can cause major headaches for legal scholars. Some claims are so obvious or obscure that they have not been made before. Other claims are made up or false, making them more difficult to support using references to the existing literature. Legal scholars need a source they can cite when confronted with these challenges. It should be something with an impressive but generic title. I offer this page, with the following conclusion: If you have been directed to this page by a citation elsewhere, it is plainly true that the author’s claim is correct.”

I found two identical packs of Skittles, among 468 packs with a total of 27,740 Skittles [PossiblyWrong]: “Why bother with nearly three months of effort to collect this data? One easy answer is that I simply found it interesting. But I think a better answer is that this seemed like a great opportunity to demonstrate the predictive power of mathematics. A few months ago, we did some calculations on a cocktail napkin, so to speak, predicting that we should be able to find a pair of identical packs of Skittles with a reasonably– and perhaps surprisingly– small amount of effort. Actually seeing that effort through to the finish line can be a vivid demonstration for students of this predictive power of what might otherwise be viewed as “merely abstract” and not concretely useful mathematics.”

Who’s Prone to Drone? A Global Time-Series Analysis of Armed Uninhabited Aerial Vehicle Proliferation [SSRN]: “What determines whether countries pursue and obtain armed drones? […] We theorize and find evidence that security threats––like terrorism––are not the only factors driving proliferation. Regime type also has a significant effect, but it varies over time. From 1994-2010 regime type had no significant effect, but non-democracies became significantly more likely to pursue and obtain armed drones from 2011-2019 due to the concurrence of three shocks in time, the most important of which asymmetrically eased supply-side constraints for non-democracies. We also find that status-seeking states are more likely to pursue armed drones. Our results contribute to the broader academic literature on proliferation by demonstrating how supply and demand shocks can lead to changes in proliferation trends over time and lending further credence to the importance of prestige in international politics.”

Why the Katy Perry/Flame lawsuit makes no sense [Youtube]: Interesting combination of music theory, law, political precidents, and good videography.

CV Dazzle sets up an awesome cyberpunk movie where all the characters have super cool looking costumes and the plot justification is “to circumvent facial recognition software”.

Maine woman stalked by drone; police unable to help - The cyberpunk future is real.

How Will You Measure Your Life?: Have a life purpose. Focus on your life purpose. It’s easy to neglect relationships but you shouldn’t. Focus on building a culture. Always stay true to your principles. Stay humble. Focus on measuring what matters.

Patrick Collison’s Advice: Go deep on a bunch of things. Work hard. Make friends over the internet with people who are great at things you’re interested in. Read a lot. Think for yourself. Don’t neglect social skills.

In Praise of Flavored Sparkling Water: “Some people have the intuition that any food or drink that is pleasurable must be bad for health. I would emphasize instead that it is important to find types of food and drink that are pleasurable enough relative to less healthy alternatives that they can help one stay away from the worst foods in the long run.”

Evidence-Based Exercise – Why and How: "Moderate and/or vigorous intensity “cardio” exercise for at least 20 minutes per day on at least 3 days per week is recommended. Additionally, resistance exercise (~ 3 x 10 repetitions with high intensity) for each of the major muscle groups should be performed on 2-3 days per week (can be combined with cardio).”

Stop trying: “If there’s one thing that obsesses self-improvement junkies, it’s actually sticking with their regimens. Life hacks abound: Tell your friends about your plan so you’ll be embarrassed if you fail! Make sure you keep an unbroken streak! Charge yourself money if you don’t stick to it! Donate the money to your least favorite charity!1 Forget donations, how about electric shocks? The escalating absurdity should tell you that this problem is unsolved. But in my opinion, that’s because everyone tries to do it wrong. Personally, I’ve found my productivity habits to be extremely long-lived when I make sure they’re literally zero maintenance. That is, when it’s habit time, it requires more willpower to break the habit than to keep going.”

100 Ways To Live Better: Some of these are probably good. I just don’t know which ones.

College-Educated Voters Are Ruining American Politics: “Instead, they are scrolling through their news feeds, keeping up on all the dramatic turns in Washington that satiate their need for an emotional connection to politics but that help them not at all learn how to be good citizens. They can recite the ins and outs of Special Counsel Robert Mueller’s investigation or fondly recall old 24-hour scandals such as Sharpiegate, but they haven’t the faintest idea how to push for what they care about in their own communities. If you think the status quo in politics isn’t great, then the time wasted on political hobbyism is pretty tragic.”

Smart People Are Not Ruining America: “We have a problem in this country. The economic elite is destroying it, and the intellectual elite is largely powerless to stop the wreckage, and while there are many sources of our powerlessness, one of the main ones is that we get the bulk of the hate. The plebeians lump us all together, because the economic elite has told them to do so. They make no distinction between the magazine columnist, who can barely afford her studio in Brooklyn, and the private-jet billionaire who just fired them by changing numbers in a spreadsheet. […] The country isn’t being destroyed by people using the word “intersectionality”. No, it’s being wrecked by the weakening of unions, corporate downsizing, accumulated environmental damage, rising anti-intellectualism, and creeping plutocracy. We have a real enemy and it’s time to put our (very mild) differences aside and fight.”

In a break from tradition, I am endorsing all 12 Democratic candidates: “In 2020, American voters must choose among several visions of the future. There is a serious debate going on in the Democratic Party about what is wrong with the country and what needs to be done to fix it. And I, for one, am not going to take a stance on that debate. Should we be realists? Should we be radicals? Should we be neither radical nor realistic? Yes, yes and yes! […] This building is either on fire or not, and consequently, I advise people either to rush out of the building with their children and valuables, or take a nap! Good luck!”

Americans are Right to Think the Economy is Rigged: “One might think that economic inequality leads to self-correction in democracies, as the public becomes alarmed or outraged by income gaps and institutes taxes or other policies to take from the rich or give to the poor. But this doesn’t happen often. Researchers have found that instead, in countries around the world, the accumulation of wealth also often leads to accumulation of political power that is then harnessed to multiply that wealth. Indeed, that’s what we’re seeing in America. Our political system responds to large donors, so politicians create benefits for the rich, who then reward the politicians who created them. How different is this from the symbiosis in the Middle Ages between a king and the nobility, elevating aristocrats who repressed the peasantry at the same time that they hailed their own magnanimity and rolled their eyes at the peasants’ morals?”

Unsafe at Any Rate: “It is impossible to buy a toaster that has a one-in-five chance of bursting into flames and burning down your house. But it is possible to refinance an existing home with a mortgage that has the same one-in-five chance of putting the family out on the street–and the mortgage won’t even carry a disclosure of that fact to the homeowner. Similarly, it’s impossible to change the price on a toaster once it has been purchased. But long after the papers have been signed, it is possible to triple the price of the credit used to finance the purchase of that appliance, even if the customer meets all the credit terms, in full and on time. Why are consumers safe when they purchase tangible consumer products with cash, but when they sign up for routine financial products like mortgages and credit cards they are left at the mercy of their creditors?”

The Choice Isn’t Between Capitalism or Socialism: “The spirited online debates about socialism and capitalism are, therefore, mostly useless. They ignore and obscure the multiple dimensions of policy, as well as changes over time, and thus make it harder rather than easier to think about concrete ways to fix the problems in the U.S. system. It would be helpful to have a new consensus terminology to describe the economic systems that various industrialized countries – the U.S., France, Japan, China, and the countries of Scandinavia – have developed over the last three decades. But one thing is for certain – the dichotomy of socialism versus capitalism, inherited from the ideological battles of the past two centuries, is badly out of date.”

Don’t Abolish Billionaires: “Fixing these policy failures might create a system that produces fewer billionaires. But that shouldn’t be the point. It might also produce more morally worthy 10-figure fortunes. That’s great, because we should be aiming to channel entrepreneurial energy into productive wealth creation that lifts us all up and away from the extraction of wealth through unjust rules that close off opportunity and deprive us of the blessings of innovation. There is a possible America where routes to extractive wealth have been closed and barriers to productive wealth have been cleared; where the wealthiest have somewhat less, and the rest of us have a great deal more. In this America, our economy and democracy are more equitable and less corrupt, and the least well-off fare better than ever. We should dearly want to live there, in a place where, if you can manage to become one, it’s more than “morally appropriate” to be a billionaire.”

A Pragmatic View on the Existence of Billionaires & Government: “The recent direction of public discourse appears to have swung in the direction that leads people to believe that billionaires or capitalists are always good or bad. These are lazy generalizations that lead to unproductive policy discussions. Yes, there are strong arguments for inequality presently and I personally believe that there are good arguments for higher capital gains taxes and personal income taxes. But we should disavow political commentaries based on vague generalizations such as “banning billionaires” or Ayn Randian style commentaries about government always being bad. And while the “both sides” police will surely accuse me of trying to see both sides, I am afraid to inform you that that’s precisely what’s needed right now because there is a lot of truth in both sides of this discussion.”

Liz Was a Diehard Conservative: “The story of Warren’s awakening—from a true believer in free markets to a business-bashing enforcer of fair markets; from a moderate Republican who occasionally missed an election to one of the most liberal senators in America vying to lead the Democratic Party—breaks the mold of the traditional White House contender and is key to understanding how she sees the world: with a willingness to change when presented with new data, and the anger of someone who trusted the system and felt betrayed.”

Explaining Warren: “This is an uncomfortable train of thought to entertain, no question. Having neglected to take adequate account of the role of power in society, we have allowed power to concentrate in ways that make reforming the system very difficult. Economic growth, economists like to say, is a positive-sum affair. If you make the right policies, then you ought to end up sufficiently rich as a society that winners from policy shifts can compensate losers and everyone ends up better off than they were before. This is the sunny picture that people on the other side of the rift—those more or less happy with the status quo—like to share with each other. It’s an appealing image! Because if no one has to lose, then everyone can get on board and no one has to feel bad. But if power is the problem, then you have a different situation. Power is necessarily zero-sum. Correcting the balance of power means reducing the power and privilege that the powerful and privileged now enjoy.”

The case for Elizabeth Warren: “The case for Warren over her competitors is threefold. She understands America’s problems better than anyone else in the field, in part because it’s her research and analysis that now forms the base for much of the policy debate. She understands how to focus and wield the powers of the regulatory state better than anyone else, because she’s actually done it, and because it’s core to her political project. And she is, far and away, the candidate with the clearest plan for making ambitious governance possible again.”

Socialists Will Never Understand Elizabeth Warren: “Warren’s politics are so confusing because we have forgotten that a pro-capitalist left is even possible. For a long time, political debate in the United States has been a fight between conservatives and libertarians on the right, who favored the market, and socialists and liberals on the left, who favored the government.”

How to argue for a wealth tax: “I played around with a bunch of numbers, and across 20-30 year periods came up with total net capital gains tax rates in the 50 to 70 percent range, noting that the current 20 percent long-term base rate for high earners is applied to nominal not real gains. Has any wealthy country sustained such a high net real capital gains rate? Of course, rhetorically a “2 percent tax on wealth” sounds much better than say “a 62 percent tax rate on long-term capital gains.” Don’t be fooled!”

The economic policy of Elizabeth Warren: Some potentially bad policies of Elizabeth Warren: She wants to ban fracking through executive order (I agree this is bad); her private equity plan would end PE which is net beneficial (I don’t know enough to say); her farm plan is too nationalistic and protectionist (I don’t know enough to say); her tax plan taxes too much (I don’t know enough to say); her student debt forgiveness and free college plans are insufficiently egalitarian, too expensive, and leads to bad incentives (I don’t know enough to say); her health care policy is too vacillating (I disagree); breaking up Big Tech would be bad (I don’t know enough to say); and she wants to do too much through executive order (I lean toward this being bad but I don’t know how else to accomplish an agenda).

Cleaning Up the Welfare State: Most conspicuously, it lacks many benefits found in similarly developed countries like child care, paid leave, and universal health care. These omissions tend to dominate the public debate, but creating a decent welfare system will require more than just adding new benefits. We also need to clean up and rationalize the welfare state we already have. The US welfare state has a lot of problems. In this paper, I propose a series of moderate reforms to major welfare programs including Medicare, Medicaid, Social Security, Supplemental Security Income, Unemployment Insurance, and Supplemental Nutritional Assistance Program. The overall purpose of these proposed reforms is to move towards a welfare state that is simpler, more coherent, and more centralized.

What Rich Centrists Could Learn from Howard Schultz about Political Power: “Wealthy centrists have an opportunity to show they’re willing to fight for democracy not through bids for the presidency in the manner of Steyer and Bloomberg, but in the way that Schultz fought in retail: with an entrepreneurial vision, a realistic plan, an army of organizers, and a willingness to compromise between values and public demands. If instead they keep to the powerless track of hobbyism, they shouldn’t be surprised when the country turns toward a form of politics they despise. And that despises them.”

The Tetris Effect: “It’s hard, often impossible, to predict the exact block that will cause the wave function collapse. The events aren’t sequential, and you often don’t know the entire board. In situations like sales and recruiting, your best bet is to work to increase the odds that you’re going to drop the right block[1] by showering your customers with goodness they’re interested in. Concretely: optimize the environment. Spend a lot of time with people you’d like to hire. Send them content they’d enjoy. Do the same with your users. Do the same with your food. Etc. Less time optimizing a specific thing, more time improving the environment around that thing.”

#3: Lessons on ‘capturing the heavens’ from the ARPA/PARC project that created the internet & PC (23 June 2017): Finding highly talented people (and only highly talented people) and having a strong vision and strong culture can lead to transformative change. Corporations naturally tend toward stasis and this trend needs to be actively resisted with an intense hostility.

Did Van Halen’s Concert Contract Require the Removal of Brown M&Ms?: “The legendary “no brown M&Ms” contract clause was indeed real, but the purported motivation for it was not. The M&Ms provision was included in Van Halen’s contracts not as an act of caprice, but because it served a practical purpose: to provide a simple way of determining whether the technical specifications of the contract had been thoroughly read and complied with. As Van Halen lead singer David Lee Roth explained in his autobiography:”

The Killer Algorithms Nobody’s Talking About: “Eventually, the lead-up to a strike may involve dozens or hundreds of separate algorithms, each with a different job, passing findings not just to human overseers but also from machine to machine. Mistakes could accrue; human judgment and machine estimations would be impossible to parse from one another; and the results could be wildly unpredictable.”

The coming technocracy: “I suspect that the trends I see in China will eventually spread elsewhere, and at some point all governments will become “potentially totalitarian” in the sense of having the technological capability of exercising total control over the population. They’ll be able to know everywhere you move, everything you read, and everything you buy. Whether they will actually exercise that control is another question, which I can’t answer. I don’t like where the world is headed. But perhaps younger people are comfortable with all this technology.”

In Favor of Niceness, Community, and Civilization: “So again we make an agreement. I won’t use the apparatus of government against Protestantism, you don’t use the apparatus of government against Catholicism. The specific American example is the First Amendment and the general case is called “liberalism”, or to be dramatic about it, “civilization 2.0”. Every case in which both sides agree to lay down their weapons and be nice to each other has corresponded to spectacular gains by both sides and a new era of human flourishing.”

The Universal Rules of Civilized Discourse: “The principles in the default Discourse community behavior FAQ were distilled, as best we could, from the common, shared community guidelines of more than 50 forums active for a decade or more. We hope that anyone can launch a Discourse-powered forum community with confidence, knowing that they’ve got the basic guidelines for civilized discourse covered.”

Human Body Temperature Has Decreased In The United States, Stanford Study Finds: “That standard of 98.6 F was established by German physician Carl Reinhold August Wunderlich in 1851. Modern studies, however, have called that number into question, suggesting that it’s too high. A recent study, for example, found the average temperature of 25,000 British patients to be 97.9 F. In a study published today in eLife, Parsonnet and her colleagues explore body temperature trends and conclude that temperature changes since the time of Wunderlich reflect a true historical pattern, rather than measurement errors or biases.”

Current Affairs’ “Some Puzzles For Libertarians”, Treated as Writing Prompts for Short Stories: “But it really only clicked a few weeks ago, when these goons from McDonalds broke into my house on a no-knock drug raid and shot my dog, and then muttered something about how surely I couldn’t object to private coercion. And it just got me thinking – what if this whole world is just a thought experiment by a communist with a crappy understanding of political philosophy trying to weak-man libertarianism? And then I thought – frick, I better get some really good vengeful-rich-person insurance, like, right away.”

Ars Longa, Vita Brevis: “Knowledge,” said the Alchemist, “is harder to transmit than anyone appreciates. One can write down the structure of a certain arch, or the tactical considerations behind a certain strategy. But above those are higher skills, skills we cannot name or appreciate. Caesar could glance at a battlefield and know precisely which lines were reliable and which were about to break. Vitruvius could see a great basilica in his mind’s eye, every wall and column snapping into place. We call this wisdom. It is not unteachable, but neither can it be taught. Do you understand?”

The Proverbial Murder Mystery: Death by proverb.

That “Away” CEO that we were previously warned about, who was previously removed, is somehow back again.

Cassette Tape Thoughts: “The packaging of intellectual positions and views is one of the most active enterprises of some of the best minds of our day. The viewer of television, the listener to radio, the reader of magazines, is presented with a whole complex of elements-all the way from ingenious rhetoric to carefully selected data and statistics-to make it easy for him to “make up his own mind” with the minimum of difficulty and effort. But the packaging is often done so effectively that the viewer, listener, or reader does not make up his own mind at all. Instead, he inserts a packaged opinion into his mind, somewhat like inserting a cassette into a cassette player. He then pushes a button and “plays back” the opinion whenever it seems appropriate to do so. He has performed acceptably without having had to think.”

Too much transparency makes the world more opaque.: “In contrast, a non-transparent, off-the-record process can reveal new information because less transparent can be more honest. The off-the-record system isn’t a guarantee of useful information, as the NYTimes has its biases and the off-the-record system only works because it is coarse, but coarse systems can reveal more information. The demand for transparency seems so innocuous. Who could be against greater transparency? But transparency is inimical to privacy. And we care about privacy in part, because we can be more honest and truthful in private than in public. A credible off-the-record system leaks a bit of honesty into the public domain and thus improves information overall. Too much transparency, in contrast, makes the world more opaque.”

Don’t Be a Modal Weasel: “I often hear academics say things like: “It is not necessarily the case that the evidence would support that.” Is this sentence meaningless or just trivial? I don’t know, but I am still surprised by how many otherwise reasonable people hide behind such verbiage. What’s the alternative? Plain, simple, clear, clean, bog-standard, unadorned probabilities. Instead of “It could be impossible,” say “The probability is 3%” or “The probability is 17%” or whatever.”

What’s the Point of Productivity?: Productivity is not work ethic or hustle. Productivity is not time management. Productivity is not just doing things that look like work. Productivity is a measure of your output divided by your input. Output is measured by the importance of the accomplishment to your goals. Input is measured by the time, energy and attention you have available.

Why Your Work Disappoints You: “It’s easier to know what other great artists/authors/etc. do to make their work great, but harder to implement that in practice. This can create a taste gap that inspires perfectionism. To get out of this rut make a schedule of regularly putting stuff out / a deadline (to avoid perfectionism) and then try to get at least 1% better each iteration at one particular part.”

4 huge costs of becoming more productive: "I didn’t actually enjoy reading The New York Times. In my head, I liked the idea of being a guy that reads The New York Times every Sunday (even though I live in Canada), so I subscribed. It’s hard to admit: I mainly read the paper so people could see me reading it. I unsubscribed shortly after I realized this, and haven’t much read much of the paper’s articles, even online, since.”

Avoiding burnout as an ambitious developer: “Be willing to say no. Know what you want and, more importantly, what you don’t want. When you look at your list of tasks that you have to accomplish in a given day, figure out how much you can realistically get done that day and actively move the rest to another day. By being realistic about what you’re going to be doing just today, you can set aside time to use the rest of your energy for the day on things that will give your brain a break. Actually make yourself a to-do list, stick to it, and don’t write so much that you always have tasks left over. By giving yourself dedicated time to rest up, you’ll be able to get your tasks done driven with more purpose, rather than driven by the looming feeling of being overwhelmed.”

Too Tired to Do Everything? How to Live Better Without Burning Out: “I write a lot about how to live better, and sometimes all those suggestions can start to feel more like burdens than helpful advice. […] Part of the problem is that suggestions are almost always given on their own. […] There have always been trade-offs. If you’re meditating for an hour every day, that might cut into your gym routine. […] If living well sounds exhausting, what should you do about it? […] First, realize that you actually cannot do everything you should. […] Second, make a distinction between a lack of inertia and persistent exhaustion from an activity being more hassle than help. […] Finally, live well for your own sake, not because somebody told you that you should.”

Intervention Profile: Ballot Initiatives: Ballot initiatives are a form of direct democracy in which citizens can gather signatures to qualify a proposed piece of legislation for the ballot, which is then subject to a binding up-or-down vote by the general electorate. Ballot initiatives are possible in Switzerland, Taiwan, many U.S. states and cities, and elsewhere. Ballot initiatives appear to maintain several advantages over more traditional policy lobbying, including lower barriers to entry and more direct control over the final legislation. However, the ultimate cost-effectiveness of a ballot initiative campaign depends on several factors, many of which are difficult to specify precisely. Although ballot initiatives hold enough promise to warrant additional investigation, it is not yet possible to say to what extent ballot initiative campaigns ought to be pursued by the effective altruism community.

EAF’s ballot initiative doubled Zurich’s development aid: The Effective Altruism Foundation (EAF), then based in Switzerland, launched and won a ballot initiative that *doubled* development aid from the city of Zurich.

The Truth About the Vaping Crisis (Freakonomics Ep. 398): Vaping is a public health win because it shifts people away from far more harmful cigarettes. However, vaping is also a public health crisis because it is targeting youth and getting people to vape who wouldn’t otherwise use any tobacco product. This crisis could be better dealt with via regulation, such as age restrictions and tobacco density restrictions. However, it seems to be dealt with outright bans over other health crises that aren’t even well connected to (tobacco) vaping.

A Lucky Stalemate: Political stalemates can be good as they tend to moderate legislation and prevent bad legislation and provide a check on corruption.

Diseased thinking: dissolving questions about disease: “People commonly debate whether social and mental conditions are real diseases. This masquerades as a medical question, but its implications are mainly social and ethical. We use the concept of disease to decide who gets sympathy, who gets blame, and who gets treatment. Instead of continuing the fruitless "disease” argument, we should address these questions directly. Taking a determinist consequentialist position allows us to do so more effectively. We should blame and stigmatize people for conditions where blame and stigma are the most useful methods for curing or preventing the condition, and we should allow patients to seek treatment whenever it is available and effective.“

Learning Day: "At OpenAI, each Thursday is Learning Day: a day where employees have the option to self-study technical skills that will make them better at their job but which aren’t being learned from daily work. We’ve found that the biggest contributions at OpenAI come from cross-functional experts, so we either need to hire them or grow them here. Before Learning Day, we very rarely saw people grow cross-functionally—for example, employees coming from a software background rarely picked up machine learning (something equally rare in other organizations except academia). Since Learning Day, this kind of growth has become very common.”

What does it mean to re-establish deterrence?: ““Reestablish deterrence” sounds firm and yes, manly. But it is backward-looking and limits the possibilities for solving the problem it is aimed at. It cannot map out a path to peace. “Deterrence” has become a get-out-of-jail-free card. It’s painted as defensive, but responding to violence with violence is escalation, not deterrence.”

Wicked Problems and The Role of Expertise and AI in Data Science: Kind environments (e.g., chess) have rigid rules, repeated patterns, and clear feedback. Wicked environments do not. All environments fall on a spectrum. Experts and AI both perform better in kind environments. Expert performance is particularly bad in wicked environments. AI can learn better through prolonged practice and is and is not prone to exhaustion, which could make for good AI-human teams.

Evaluating State and Local Business Tax Incentives: That thing where states offer companies tax breaks to move their business into the state don’t provide good returns.

Hiring the Top 1%: “I’m exaggerating a lot, but the point is, when you select 1 out of 200 applicants, the other 199 don’t give up and go into plumbing (although I wish they would… plumbers are impossible to find). They apply again somewhere else, and contribute to some other employer’s self-delusions about how selective they are.

The Dark Secrets of BERT: (1) BERT is heavily overparametrized. This means many weights, and even entire layers, can be dropped without loss in performance. This also explains why BERT can be so easily distilled. (2) BERT does not need to be all that smart for these tasks. BERT can perform fairly well without pre-training (though does best with pre-training). This also means that instead of verbal reasoning, BERT may learn to rely on various shortcuts, biases and artifacts in the datasets to arrive at the correct prediction. It also means that BERT’s success may be due to something other than self-attention.

Does GPT-2 Understand Anything?: "In summary, it seems that GPT-2 does have something that can reasonably be called “understanding” and holds something very much like “concepts” or “ideas” which it uses to generate sentences. However, there are some profound differences between how a human holds and uses ideas and how GPT-2 does, which are important to keep in mind.”

Physicist Wins Ig Noble Prize For Study On Whether Cats Should Be Classified As Liquids Or Solids: “In his official research paper, Fardin discusses many factors including relaxation time, experimental time, the type of container, and the cat’s degree of stress. The conclusion? Cats can be either liquid or solid, depending on the circumstances.”

Five More Things Millennials Need to Kill: snail mail, TV, long commutes, car ownership, and sightseeing

Are We the Baddies?: It’s a skull!

I want to put my understanding of how the world works by making a bunch of judgements about things. I doubt I’ll be right on everything, but my hope is to at least be well calibrated (that is, the things I say will happen 40% of the time happen about 40% of the time). Also, being wrong gives me a good concrete chance to see if something I think about the world is wrong and adjust my strategy accordingly.

So far, my strategy is basically to try to assume the status quo will hold, trust the right experts, and be careful of straying away from the consensus unless I feel like I really know what I’m doing. Let’s see how it goes!

Current predictions were made on 14 Jan 2020. I might edit to add more predictions if I think of them. Most questions so far were adapted from Vox’s Future Perfect article, David Manheim’s article, PredictIt (a prediction betting market), Metaculus (a community forecasting site), PredictionBook (another community forecasting site), and the Good Judgment Open (a third community forecasting site).

Donald Trump wins the US Election in 2020 - 40%

Fundamentals models favor Trump, but have a poor track record. Base rate for incumbent re-election is ~55%. The Metaculus consensus as of 14 Jan 2020 is a median of 46%. Dylan Matthews (Vox) picks 55%, mainly citing fundamentals models. David Manheim picks 40%. ElectionBettingOdds has Trump at 54% as of 14 Jan 2020.

I am skewing low due to Trump’s high degree of scandal and unpopularity. I would probably go 45%, but I’m adjusting down further to account for the fact that Trump might also get removed from office or resign early (or die - he’s old).

Update 15 Jan 4:20pm ET - MC thinks Trump is 45% likely to win if Bernie is the Dem nominee and that Trump is 65% likely to win if Biden is the Dem nominee.

Update 19 Jan 6pm ET - KL thinks 45%

After the Democratic nominee is chosen, I will still assign between 35% and 45% to the chance of Trump winning the general election - 45%

(Added 15 Jan at 4:09pm ET)

That is, I don’t expect the choice of nominee or events between now and the end of the primary to matter much.

I put this on PredictionBook.

Update 15 Jan 6:30pm ET - I forgot that by the time of the nomination there should be reasonably decent polling that could be predictive, as well as potentially a 538 model. This could shape my view a lot. Accordingly, I now drop confidence in this proposition to 50% (from 80%).

Update 20 Jan 2:30pm ET - Based on the model in “Electability and the Senate in 2020″ I have shifted my prediction from to 40% (from 50%), thinking candidate quality might matter more than I thought.

Update 20 Jan 3:30pm ET - I have shifted my prediction up to 45% (from 40%) after looking at the spread of conditional probabilities in betting markets.

The Senate removes Trump from office sometime in 2020 - 5%

The Metaculus consensus as of 14 Jan 2020 is a median of 9%. PredictIt has 10% odds. David Manheim picks 2%. I will go with 5%, balancing the fact that this seems basically impossible to happen with the possibility of some unknown unknown event.

Update 19 Jan 6pm ET - KL thinks 30%

The Senate either acquits or removes Trump from office in 2020 or Trump resigns from office in 2020 (Basically, the trial happens in 2020) - 95%

David Manheim picks ~87%. I will go with 90%. See this closely related PredictionBook.

Update 19 Jan 6pm ET - KL thinks 98%

Update 20 Jan at 12:30pm ET - Changed my prediction to 95% (from 90%) based on the fact that the trial now seems definitely underway.

Update 5 Feb at 5pm ET - This resolves correct.

The Senate trial for Trump’s impeachment will have at least one witness give testimony (public or private) in 2020 - 72%

(Added 18 Jan at 4pm ET)

The current Republican talking points are to not have witnesses, but I think there’s a lot of public pressure to have witnesses, even though the Republicans have control. I adjust downward a bit to account a bit for my above chance that the trial doesn’t happen.

I put this on PredictionBook.

Update 19 Jan 6pm ET - KL thinks 20%

Update 20 Jan at 12:30pm ET Changed my prediction to 72% (from 57%) based on a better understanding of relevant Senate rules.

Update 5 Feb at 5pm ET - This resolves incorrect.

Conditional on Trump’s impeachment trial having at least one witness give testimony, Hunter Biden will give testimony - 60%

(Added 18 Jan at 4pm ET)

Another Republican talking point is that if there are witnesses at the trial, we should hear from Hunter Biden. I could imagine this happening as some sort of “compromise”. Note that this prediction will not be scored (resolve ambiguous) if Trump’s impeachment trial does not have at least one witness give testimony.

I put this on PredictionBook.

Update 19 Jan 6pm ET - KL thinks 15%

Update 5 Feb at 5pm ET - The conditional did not happen, so this resolves ambiguously.

The Democratic Nominee will be Joe Biden - 47%

The Metaculus consensus as of 14 Jan 2020 is a median of 44%. Kelsey Piper (Vox) picks 60%, citing him as a frontrunner in the polls. David Manheim picks 40%. ElectionBettingOdds has Joe Biden at 31% as of 14 Jan 2020. The 538 model, as of 14 Jan 2020, has Joe Biden at 46% odds to win a plurality of delegates and 40% odds to win a majority of delegates. I don’t think I have any insight beyond the 538 model, so I will split the difference at 45% 47% to make for a contested convention adjustment.

Update 15 Jan 2020 at 6pm ET - Moved this from 45% to 47% to make my presidential predictions sum to 100%.

Update 15 Jan 2020 at 6:30pm ET - MS thinks Biden is 40% likely to win.

Update 19 Jan 6pm ET - KL thinks 35%

The Democratic Nominee will be Bernie Sanders - 24%

The Metaculus consensus as of 14 Jan 2020 is a median of 25%. David Manheim picks 20%. ElectionBettingOdds has Bernie Sanders at 29% as of 14 Jan 2020. The 538 model, as of 14 Jan 2020, has Bernie Sanders at 26% odds to win a plurality of delegates and 22% odds to win a majority of delegates. I still don’t think I have any insight beyond the 538 model, so I will split the difference at 23% 24%.

Update 15 Jan 4:09PM ET - MC thinks Bernie Sanders is 35% likely to win.

Update 15 Jan 2020 at 6pm ET - Moved this from 23% to 24% to make my presidential predictions sum to 100%.

Update 15 Jan 6:30PM ET - MS thinks Bernie Sanders is 30% likely to win.

Update 19 Jan 6pm ET - KL thinks 20%

The Democratic Nominee will be Elizabeth Warren - 15%

The Metaculus consensus as of 14 Jan 2020 is a median of 14%. David Manheim picks 15%. ElectionBettingOdds has Elizabeth Warren at 12% as of 14 Jan 2020. The 538 model, as of 14 Jan 2020, has Elizabeth Warren at 16% odds to win a plurality of delegates and 13% odds to win a majority of delegates. I still don’t think I have any insight beyond the 538 model, so I will split the difference at 14% 15%.

Update 15 Jan 2020 at 6pm ET - Moved this from 14% to 15% to make my presidential predictions sum to 100%.

Update 19 Jan 6pm ET - KL thinks 35%

The Democratic Nominee will be Pete Buttigieg - 10%

(Added 15 Jan at 6pm ET)

ElectionBettingOdds has Buttigieg at 6% as of 15 Jan 2020. The 538 model, as of 14 Jan 2020, has Buttigieg at 10% odds to win a plurality of delegates and 8% odds to win a majority of delegates. I still don’t think I have any insight beyond the 538 model, so I will split the difference at 10%. Also on PredictionBook.

Update 19 Jan 6pm ET - KL thinks 7%

The Democratic Nominee will be Andrew Yang - 1%

(Added 15 Jan at 6pm ET)

ElectionBettingOdds has Yang at 2% as of 15 Jan 2020. The Metaculus consensus is a median of 1% as of 15 Jan 2020. The 538 model, as of 15 Jan 2020, has someone who is not Biden, Sanders, Warren, or Buttigieg with odds to win a plurality of delegates at 2% and odds to win a majority of delegates at 1%. I still don’t think I have any insight beyond the 538 model, so I will go with 1% for Yang.

Update 19 Jan 6pm ET - KL thinks 3%

The Democratic Nominee will be Amy Klobuchar - 1%

(Added 15 Jan at 6pm ET)

ElectionBettingOdds has Klobuchar at 1% as of 15 Jan 2020. My logic to this is pretty similar to my logic for Yang.

I put this on PredictionBook.

Update 19 Jan 6pm ET - KL thinks 0%

The Democratic Nominee will be Mike Bloomberg - 1%

(Added 15 Jan at 6pm ET)

ElectionBettingOdds has Bloomberg at 11% as of 15 Jan 2020, which is a lot higher than the 538 model, which presumably puts Bloomberg at <1%. My logic to this is pretty similar to my logic for Yang and Klobuchar. I’m putting a bit more weight on Bloomberg than 538 would, though, because of the fact that Bloomberg’s large money spending - which I think will ultimately be ineffectual - may end up being an important source of model uncertainty.

I put this on PredictionBook.

Update 19 Jan 6pm ET - KL thinks 0%

The Democratic Nominee will be someone other than Biden, Sanders, Warren, Buttigieg, Klobuchar, Yang, or Bloomberg - 1%

(Added 15 Jan at 6pm ET)

This is just mathematically implied by my other predictions.  ElectionBettingOdds has this at 6.8% as of 15 Jan 2020. If this happened, something pretty weird and unprecedented would have to happen that has never happened before in US politics in the modern primary system (1970-present).

I put this on PredictionBook.

Update 19 Jan 6pm ET - KL thinks 0%

The Republican Nominee will be Donald Trump - 90%

There isn’t really an analogue for this question in other places, but I assume it is pretty similar to whether Trump completes his first term, which, as of 14 Jan 2020, has a Metaculus consensus of a median of 90% and ElectionBettingOdds has it at 91%. I will also put it at 90%.

Update 19 Jan 6pm ET - KL thinks 95%

Conditional on Donald Trump losing the election in 2020, he will contest the election results and not concede for  seven days or longer - 10%

(Added 18 Jan at 4pm ET)

Donald Trump has previously claimed that the 2016 Presidential election against him was “rigged” (though he won anyway). Trump has also made claims of widespread voting fraud in the 2016 election. Will he repeat these claims again and back it up by not conceding? Doing so would be unprecedented, but I’m thinking this is pretty unlikely, so I will go with 10%.

Note that this prediction will not be scored (resolve ambiguous) if Trump wins the Presidential election in 2020.

I added this to Metaculus and it is pending.

Update 19 Jan 6pm ET - KL thinks 99%

Update 5 Feb at 5pm ET - The metaculus consensus is now a median of 55%.

The candidate who wins the electoral vote in the US 2020 Election will also win the popular vote - 89%

(Added 18 Jan at 5pm ET)

The odds of this happening in the 2016 election was 89% according to the 538 model at the time and I don’t think the fundamentals have changed all that much. As of 18 Jan at 5pm ET, The Good Judgement Open gives this a 85% chance.

Update 19 Jan 6pm ET - KL thinks 40%

Tulsi Gabbard will drop out before the Hawaii primary - 70%

Predicting that Gabbard won’t win is easy – the 538 polling average has her at 1.6% nationally and the 538 model has her with a <0.1% chance of winning Iowa. However, it’s harder to predict how long a candidate will try to remain in the race and many candidates continue even after it is very clear they won’t win. The Hawaii Primary is 4 April 2020. The 2016 Republican Primary only had three candidates stick it out that long, which suggests a rather high base rate for dropping out earlier. On the other hand, Gabbard is not running again for Congress so she doesn’t have that much else to do as long as she can continue to raise enough money to operate her campaign.

As of 14 Jan 2020, the Metaculus consensus is a median of 50%. I’m going to average this with the base rate and conclude a 70% chance she drops out early.

Update 19 Jan 6pm ET - KL thinks 80%

The Democratic primary will be essentially settled after Super Tuesday (one candidate hits 90% or higher in prediction markets by March 5) - 50%

Kelsey Piper (Vox) picks 60%. David Manheim picks 45%. As of 14 Jan 2020, Metaculus has no consensus yet.

I also keep in mind that the prediction markets tend to be quite swingy, having already once placed Elizabeth Warren over 50% odds to win in 2019 (though of course she’s not there now).

If you had insider access to the 538 Forecast model you should be able to derive their odds too, but it seems to imply a ~30% odds that Biden wins a sizable majority of delegates on Super Tuesday (different though than being enough for a 90% victory). Additionally, if Biden wins Iowa, which he is ~30% likely to do, I think that could be enough to hit 90%. Lastly, if Sanders wins both Iowa and New Hampshire, which he is ~20% likely to do, I think that could be enough to hit 90%. Based on this, I’d go with 50%.

Update 17 Jan at 7:30am ET - The Metaculus consensus is a median of 33%.

Update 19 Jan 6pm ET - KL thinks 70%

A candidate other than Biden, Buttigieg, Sanders, Warren, or Yang wins at least one state in the Democratic Primary  - 14%

As of 14 Jan 2020, The 538 model has a ~2.5% chance that someone other than these candidates wins Iowa (mostly Klobuchar) and gives Steyer a 2% chance of winning in South Carolina. Based on this, I think a ~5% overall chance seems plausible. Also see this PredictionBook.

Update 15 Jan 2020 at 6pm ET - I forgot that Amy Klobuchar has a good chance of winning Minnesota if she is still in the race then. 538 puts this at 11%. Based on this, I have raised my guess as to this happening to 14% (from 5%).

Update 19 Jan 6pm ET - KL thinks 90%

A third-party candidate gets more than 5% of the vote in the 2020 US Presidential Election - 10%

David Manheim picks 15%. Since 1980 (10 presidential elections), this happened three times (Anderson in 1980, Ross Perot in 1992, and Ross Perot in 1996), making for a base rate of ~30%. I assume Perot was a bit of an outlier though, and I doubt other Perot-equivalents (e.g, Bloomberg) appear interested. I’d argue that the climate is ripe for it like Manheim does, but that failed to happen in 2016 where the climate seemed ripe-est. Therefore I will bet a bit against Manheim (and the base rate) and go for 10%.

I added this to Metaculus and it is pending.

Update 19 Jan 6pm ET - KL thinks 10%

The GOP holds the Senate - 65%

Dylan Matthews (Vox) picks 80%. David Manheim picks 55%. ElectionBettingOdds has 71% as of 14 Jan 2020. There is no Metaculus consensus yet, as of 14 Jan 2020.

The Senate currently has 47 Democrat-affiliated Senators and 53 Republican-affiliated Senators. In order for the GOP to not hold the Senate, the Democrats must either (a) gain a net of +4 Senate seats or (b) gain a net of +3 Senate seats and the US Presidency.

To get some probabilities, there is no 538 model yet but you can get an initial look from Inside Elections, the Cook Political Report, and Sabato’s Crystal Ball. Of course, two problems remain with aggregating these probabilities: (1) they aren’t independent events and there will be a strong intercorrelation between states and (2) there is also a strong intercorrelation between the Senate result in a state and the Presidency result in that state. Luckily for me, though, I am just making a binary forecast rather than trying to guess the number of seats won, so I can compensate by being wrong in some areas by still being directionally right.

Thus it looks like the Democrats will most likely have to (a) pick up Colorado, Arizona, and one of North Carolina or Maine and (b) hold both Alabama and Michigan… with a decent chance that something weird could happen on both sides with Virginia, New Mexico, New Hampshire, Minnesota, Iowa, Kansas, Georgia, Texas, Kentucky, or Montana. These things sound _quite _hard. But at the same time, the Democrats are likelier than not to win the Presidency, and that is also correlates with winning the Senate pretty well… so 70% sounds right.

Update 15 Jan at 4:20pm ET - The Metaculus consensus now has a median of 72%.

Update 15 Jan at 6:30pm ET - MS thinks this is 64% likely.

Update 19 Jan 6pm ET - KL thinks 75%

Update 20 Jan at 2:30pm ET - Based on the model in “Electability and the Senate in 2020″ and MS’s prediction, I have revised my prediction to 65% (from 70%).

The Democrats hold the House - 70%

ElectionBettingOdds has 70.3% as of 14 Jan 2020. In 2016, the 538 model had a ~75% chance of the Democrats winning the House in August 2018, when the congressional ballot was ~+8% for Democrats. Right now, the congressional ballot is ~+6%, which would imply lower than ~75% odds, though I’m not sure how much lower. I think I will also go with 70%.

I added this to Metaculus and it is pending.

Update 19 Jan 6pm ET - KL thinks 85%

Democrats win both halves of Congress + Presidency - 25%

David Manheim picks 40%. If I have the Democrats at 30% to win the Senate, 70% to keep the house, and 60% to win the Presidency – and these are all pretty correlated – I will go with 25%.

I added this to Metaculus and it is pending.

Update 15 Jan at 6:30pm ET - MS thinks this is 30% likely.

Update 19 Jan 6pm ET - KL thinks 65%

States (+DC) adopting the National Popular Vote Interstate Compact achieve 270 combined electoral votes in 2020 - 1%

The NPVIC currently has 196 electoral votes, but has an increasingly tough ride toward getting more states on board. In fact, it may now start going backwards. It doesn’t look like there are enough states considering passing NPVIC legislation to make this politically possible even if it were popular (since there is a lag time in considering legislation), which makes this essentially impossible, so I will go with 1%.

Update 19 Jan 6pm ET - KL thinks 25%

The states (+DC) adopting the National Popular Vote Interstate Compact have less combined electoral votes at the end of 2020 than they did at the end of 2019 - 35%

Speaking of which, the Colorado (9 electoral votes) repeal looks competitive. The Nevada (6 electoral votes) passing is dead. I imagine Maine (4 electoral votes) is less likely to pass than not, whereas I imagine Oregon (7 electoral votes) is likelier to pass than not. Based on this, to not lose more electoral votes it would require either (a) no Colorado repeal or (b) Maine and Oregon sign on. Based on that, I think 35% odds sound right.

I added this to PredictionBook.

Update 19 Jan 6pm ET - KL thinks 5%

The Supreme Court will affirm (decide in favor of abortion restrictions and overturn Whole Woman’s Health) in June Medical Services LLC v. Gee - 80%

As of 14 Jan 2020, there is no Metaculus consensus yet. Dylan Matthews (Vox) picks 90%. David Manheim picks 97.5% 87.5%, citing FantasySCOTUS. However, as of 14 January 2020, FantasySCOTUS seems to place very high confidence on reversing the ruling – so either it changed, Manheim is reading it backwards, or I am reading it backwards. However, it also looks like the FantasySCOTUS is based on only one user prediction, so I’m not sure how much stock to put into that. (Update 16 Jan 8am ET - Manheim agrees with me and changes his prediction from 97.5% to 87.5%)

I’m pretty uninformed about this so I think my prior before seeing anyone else’s predictions would be 70%. I think I would move to 85% based on hearing Matthews’s and Manheim’s predictions, and then move to 80% based looking more into the FantasySCOTUS take.

Update 15 Jan at 5:30pm ET - To be clear, this means “affirm”.

Update 17 Jan at 8am ET - The Metaculus consensus is 65%.

Update 19 Jan 6pm ET - KL thinks 45%

Trump does not get another chance to nominate a Supreme Court justice in 2020 (regardless of whether the nomination happens or succeeds) - 80%

Dylan Matthews (Vox) picks 70%, mainly citing actuarial tables. David Manheim picks 85%, mainly citing that it is silly to look at raw actuarial tables without also looking at socioeconomic status. One closely related (but not identical) Metaculus question has no consensus yet as of 14 Jan 2020. There is another closely related Metaculus question, but it is for the entire Presidency of Donald Trump, which includes a possible re-election. Since 1990 there have been ten vacancies, implying a base rate for this question being 67%. Based on this I will go for 80%.

Update 15 Jan 4:20pm ET to point out that both Metaculus questions are similar but not identical. I put 85% on the first Metaculus question and 70% on the second Metaculus question. The first question has a Metculus consensus now (70% median) and the second question is at 55% median and 32% by the fancy algorithm I don’t understand.

Update 18 Jan 5:20pm ET - This is 75% on the Good Judgment Open.

Provided new numbers on global poverty are provided between now and the end of January 2021, it will show that the absolute number of people in global poverty will fall relative to 2015 figures - 65%

Dylan Matthews (Vox) picks 60%. David Manheim picks 70%. As of 14 Jan 2020, Metaculus has no consensus yet. I don’t really have any other basis to make a prediction so I will go with 65%.

Updated on 15 Jan 4:20pm ET to more clearly match the Metaculus phrasing. There still is no consensus on the Metaculus question.

Update on 17 Jan 8am ET - The Metaculus consensus is a median of 90%.

Update 19 Jan 6pm ET - KL thinks 99%

Brexit happens by the end of 2020 - 95%

Kelsey Piper (Vox) picks 95%, citing the large Conservative victory in the UK elections. David Manheim picks 90%. ElectionBettingOdds has 98% as of 14 Jan 2020. As of 14 Jan 2020, there is no Metaculus consensus yet. 95% sounds about right.

Update on 15 Jan 4:20pm ET - the Metaculus consensus is now a median of 90%.

Update 19 Jan 6pm ET - KL thinks 90%

Update 31 Jan at 11pm ET - This resolves correct.

There are no US ground troops in Iran in 2020 - 90%

Dylan Matthews (Vox) picks 80%. David Manheim picks 85%.

One closely related Metaculus question has a consensus as of 14 Jan 2020 of a median of 5%. Another closely related Metaculus question has a consensus as of 14 Jan 2020 is a median of 10%. Neither of these two questions seem to aim solely for ground troops alone.

The recent base rate of the US sending ground troops to another country is also probably around 10% or so.

While there has been a dust up around Iran, I expect the past wars in Iraq to make war in Iran politically unpopular absent some clearly provoked attack by Iran, which I expect them to not do. Based on all of this, I will pick 90%.

Update 19 Jan 6pm ET - KL thinks 95%

No nuclear weapon used in 2020 for offensive military purposes (not a test) with at least one death - 99%

David Manheim picks 85% likely to not be used by 2025, though I don’t know what that would be over a one-year time frame (I’m trying to keep my predictions near-term for now) and I’m not sure exactly what he means by “used”. Other data collected by Rethink Priorities researcher Luisa Rodriguez points to a roughly 1% average per year chance of nuclear weapon use, which I will back.

Update 16 Jan 8:30am ET to clarify this means “No nuclear weapon used in 2020 for offensive military purposes (not a test) with at least one death” not the more vague “No nuclear weapon used in 2020″.

Update 19 Jan 6pm ET - KL thinks 99%

Update 20 Jam 10am ET - Looks like it is me vs. the millennials on this one.

China’s internment camps for Muslims will remain open - 85%

Sigal Samuel (Vox) picks 85%. David Manheim agrees. As of 14 Jan 2020, there is no Metaculus consensus yet. I don’t have any reason to buck this consensus.

Update 17 Jan at 8am ET - The Metaculus consensus is a median of 85%.

Update 19 Jan 6pm ET - KL thinks 99%

Netanyahu will not be unseated as Israeli prime minister at all in 2020 - 50%

Sigal Samuel (Vox) picks 55%. David Manheim picks 52.5%. As of 14 Jan 2020, there is no Metaculus consensus yet and PredictIt gives ~34%. I will go with 50%.

Update 17 Jan at 8am ET - The Metaculus consensus is a median of 40%.

Update 18 Jan at 5pm ET - The Good Judgment Open consensus is 58%.

Update 19 Jan 6pm ET - KL thinks 5%

Before 13 July 2020, the United States and Afghan Taliban will sign a peace agreement - 40%

(Added 22 Jan at 9am ET)

The Good Judgement Open gives 25%.

Maduro will still be President of Venezuela by the end of 2020 - 80%

As of 14 Jan 2020, PredictIt gives ~76%. I will go with 80%. See on PredictionBook.

Update 18 Jan 2020: The Good Judgement Open gives 96%.

Update 19 Jan 6pm ET - KL thinks 60%

Scotland will announce an independence referendum in 2020 - 5%

As of 14 Jan 2020, PredictIt gives ~32%. I will go with 30%.

I added this to PredictionBook.

Update 18 Jan 2020 at 5pm ET: The Good Judgement Open forecast is 5%. I will also change my forecast to 5%.

Update 19 Jan 6pm ET - KL thinks 35%

The US-China trade war still be ongoing by November 2nd, 2020 - 40%

As of 14 Jan 2020, the Metaculus consensus is a median of 44%. I think betting on the status quo (trade war continuing) is a good prior. There certainly have been attempts to cool the trade war so far, but they haven’t panned out yet. That being said, there is increasing economic pressure and Trump may want to declare some sort of “victory” on the trade war prior to the US election. Latest news also continues to point ambiguously but toward some sort of resolution. I will go with 40%.

Update 19 Jan 6pm ET - KL thinks 45%

The S&P 500 will end at a higher price on 31 Dec 2020 as it did on 31 Dec 2019 - 60%

(Added 18 Jan at 2pm ET)

Stocks usually go up. Since 1927, the S&P 500 has closed higher 61 times and lower 31 times, for a base rate of 66%. Of the 60 times the S&P 500 closed the previous year higher, it also closed higher again the following year 39 times, for a conditional base rate of 65% - not much of a difference. On the other hand, there are some signs that we’ve reached a high point and it is hard to go higher. On the other other hand, the stock market is basically impossible to predict beyond base rates. As of 18 Jan, the Metaculus consensus is a median of 53%, but I will err closer to the base rate and go with 60%.

Update 19 Jan 6pm ET - KL thinks 85%

Trump meets with Kim in 2020 - 30%

As of 14 Jan 2020, PredictIt has 33%. The base rate seems to be one out of three years (2018 out of a possible 2017, 2018, and 2019). I will go with 30%.

I put this on PredictionBook.

Update 18 Jan 2020 at 5:20pm ET: This is 14% on Good Judgment Open.

Update 19 Jan 6pm ET - KL thinks 55%

The US announces a withdrawal from NATO in 2020 - 5%

As of 14 Jan 2020, PredictIt has 6%. ~5% sounds about right. This is crazy but Trump is also crazy, so I guess it could happen. See this on PredictionBook.

Update 19 Jan 6pm ET - KL thinks 45%

No gene drives to fight malaria-carrying mosquitoes will be launched in any part of the world - 85%

Dylan Matthews (Vox) picks 90%. David Manheim picks 85%. As of 14 Jan 2020, there is no Metaculus consensus yet. I will also go with 85%.

Update 17 Jan at 8am ET - The Metaculus consensus is 85%.

No new CRISPR-edited babies will be born - 80%

Kelsey Piper (Vox) picks 80%. David Manheim picks 85%. As of 14 Jan 2020, there is no Metaculus consensus yet. I will also go with 80%.

Update 15 Jan at 8am ET - The Metaculus consensus is 80%.

Update 19 Jan 6pm ET - KL thinks 30%

The number of drug-resistant infections will increase - 85%

Sigal Samuel (Vox) picks 70%. David Manheim picks 85%. As of 14 Jan 2020, this does not appear to be on Metaculus. I will also go with 85%.

Update 19 Jan 6pm ET - KL thinks 99%

There will be more than 8000 human infections of the 2019 Wuhan novel coronavirus (2019-nCoV) reported before 2021 - 15%

(Added 22 Jan at 9am ET)

The SARS-CoV outbreak in 2003 resulted 8000 infected and killed nearly 800 people. Will the 2019-nCoV exceed that? My rather strong inclination is no, as it seems like we have learned a lot from SARS-CoV and are taking it much more seriously. There is a related Metaculus on this but it is hard to compare.

See also this Prediction Book.

Update 31 Jan at 10am PT - This resolves correct. In retrospect, I think I was overconfident here.

Facial recognition will be banned in at least three more US cities - 80%

Sigal Samuel (Vox) picks 70%. David Manheim picks 80%, citing it as an easy thing for cities to do. As of 14 Jan 2020, the Metaculus consensus is a median of 70%. I will also go with 80% - there are a lot of cities out there.

Update 19 Jan 6pm ET - KL thinks 95%

Update 23 Jan 7pm ET - now banned in Cambridge MA

Beyond Meat will outperform the general stock market - 55%

Kelsey Piper (Vox) picks 70%. David Manheim picks 55%, citing stock market efficiency hypothesis. As of 14 Jan 2020, this does not appear to be on Metaculus. Presumably “general stock market” is defined as S&P 500. I also agree with the market efficiency hypothesis and assume all the rosiness for Beyond Meat is already priced in, and go with 55%.

Update 19 Jan 6pm ET - KL thinks 90%

Update 19 Jan 6pm ET - This is upcoming on Metaculus.

Global carbon emissions will increase relative to 2019 - 85%

Kelsey Piper (Vox) picks 80%. David Manheim picks 85%. As of 14 Jan 2020, there is no Metaculus consensus yet. I will also go with 85%.

Update 15 Jan at 4:20pm ET - The Metaculus consensus is a median of 85%.

Update 19 Jan 6pm ET - KL thinks 99%

Average world temperatures will increase relative to 2019 - 66%

Sigal Samuel (Vox) picks 60%, citing an increased trend toward hotter climate offset by occasional shocks like the 2016 El Niño. David Manheim picks 65%. As of 14 Jan 2020, the Metaculus consensus is a median of 66%. I will also go with 66%.

Update 19 Jan 6pm ET - KL thinks 99%

California has a wildfire among the 10 most destructive in state history - 40%

Kelsey Piper (Vox) picks 60%. David Manheim picks 40%, arguing that Piper is overweighing recent events and not counting for a strongly increased safety response. As of 14 Jan 2020, there is no Metaculus consensus yet. I will also go with 40%.

Update 17 Jan at 8am ET - The Metaculus consensus is a median of 59%.

Update 19 Jan 6pm ET - KL thinks 65%

California has an earthquake with either one death or >$1B in damage in 2020 - 18%

(Added 18 Jan at 4pm ET)

This list of earthquakes in California finds 14 years with such earthquakes since 1940 (’52, ‘54, ‘57, ‘69, ‘71, ‘87, ‘89, ‘91, ‘92, ‘94, ‘03, ‘10, ‘14, ‘19), for a base rate of 14/79 = 18%. I don’t know how to improve on that base rate, so I will go with 18%. This base rate could be different if you chose a different cut-off year.

I put this on PredictionBook.

Tesla has at least one truly autonomous (no human safety driver) taxi (where a member of the public chooses the destination and pays) on a regular street (where unrestricted human-driven cars are driven on the same street) by the end of 2020 - 5%

Taken from this Rodney Brooks tweet (though he doesn’t state odds). See also “AGI Has Been Delayed”.

I put this on PredictionBook.

Update 19 Jan 6pm ET - KL thinks 5%

Bitcoin ends the year below $7500 - 40%

As of 14 Jan 2020, Bitcoin is around $8,800. It hasn’t been below $7500 since May 2019. My prior is it is about as likely to go up as it is to go down (efficient market hypothesis) and I don’t have any information about bitcoin to really suggest otherwise. Given that it is already up, I err on the side of bitcoin. See this on PredictionBook.

Update 15 Jan 2020 at 7pm ET - I revised this down to 40% (from 45%) after reconsidering a bit more, but my reasoning is unchanged.

Update 19 Jan 6pm ET - KL thinks 45%

Facebook’s Libra will launch by 1 October 2020 - 7%

(Added 22 Jan at 9am ET)

It seems like the moment is gone and momentum is massively in reverse, so I give this 7%. As of 22 Jan 2020, the Metaculus consensus is a median of 17%.

NASA will not successfully recontact the ASTERIA satellite before the end of March 2020 - 89%

(Added 18 Jan at 2:20pm)

This question came up on Metaculus and as of 18 Jan 2020 has no consensus yet and no other predictions for me to anchor on, so I thought I’d try my hand at predicting it. Popular Mechanics says “there’s a small likelihood they could still reestablish contact”… how small?

The satellite has been lost since 5 Dec 2019. Looking at some other examples of lost satellites, I don’t see other examples of satellites being lost for 1.5 months that have been recontacted, but my sample size is only seven (IMAGE x2, ISEE-3, STEREO-B, WALL-E, EVE, Philae).

The article “If nothing goes wrong, is everything all right? Interpreting zero numerators” (Hanley & Lippman-Hand, 1983) gives a procedure for converting 0/7 into an upper bound 95% confidence interval of no more than 3/7 = 43%.

Using Monte Carlo simulation, we would observe 0/7 about 50% of the time if the initial odds were ~11%.

Laplace’s Rule of Succession states that “If we repeat an experiment that we know can result in a success or failure, n times independently, and get s successes, and n-s failures, the odds of a success on the next run are (s+1)/(n+2)”… so (0+1)/(7+2) = 1/9 = ~11%.

Note that this assumes a uniform Bernouli prior (everything is a priori equally likely)… One could adapt the rule to (s+a)/(n+b) where a/b is the prior.

Update 19 Jan 6pm ET - I recorded this prediction backwards, forgetting the negation. I meant to say 89% when I said 11%.

Update 19 Jan 6pm ET - KL thinks 90%

Update 19 Jan 6pm ET - The Metaculus consensus is a median of 80%.

Greta Thunberg will win the Nobel Peace Prize in 2020 - 10%

I think if this was going to happen, it would’ve happened in 2019. Also the base rate of one particular person winning the Nobel Peace Prize is quite low. So I will go with 10%. See this on PredictionBook.

Update 19 Jan 6pm ET - KL thinks 75%

The Doomsday Clock will advance closer to midnight in 2020 - 35%

Since January 2018, the clock has been set at two minutes to midnight. The clock has been set 26 times, of which 15 were advancements for a base rate of ~58%. However, the last time it was this close, the US and the USSR had just tested their first thermonuclear weapons and it’s quite hard to argue that things are even worse now, except maybe with regard to climate change. On the other hand, not advancing it at least a little bit might harm their credibility. As of 17 Jan, the Metaculus consensus is a median of 37%. I will go with 35%.

Update 19 Jan 6pm ET - KL thinks 95%

Update 23 Jan 2020 @ 12PM ET: This resolved correct.

There will be a 4x GM in Kaggle by the end of 2020 - 20%

(Added 15 Jan at 10:30pm ET)

I imagine you won’t really know what this means if you’re not on Kaggle (an international data science competition), but I was asked to predict this, so here it goes.

Abhishek is the closest to 4x GM by a large margin. He already is a 3x GM and the category for which he doesn’t have GM status - datasets - he is still #2 ranked. However, he still needs 5 gold dataset medals. Chris Crawford is the only dataset GM and it looks like it took him ~2 years to do it. Additionally, datasets have a strong “low hanging fruit” problem as really popular widely known (and thus widely upvoted) datasets are only added once, thus making it hard to repeat Crawford’s success. Looking at some other dataset profiles, it does look hard to get 5 gold dataset medals in one year. Based on this, I will put 20% odds that there will be a 4x GM in 2020.

I put this on PredictionBook.

Ken Jennings wins the Jeopardy Greatest of All Time Tournament - 70%

As of 14 Jan 2020 at 6:58PM ET, Ken Jennings is up 2-1 games over James Holhauzer (1 win) and Brad Rutter (0 wins) in a “best of seven” tournament. Based on how play has unfolded so far, I think Jennings is 70% likely to win. (I made this timestamped non-editable prediction in advance of tonight’s 14 Jan game as proof. Also note that the publication date of this post is before that of the game.)

Update 14 Jan 2020 @ 9:01PM: This resolved correct.

Ken Jennings wins the Jeopardy Greatest of All Time Tournament tonight - 45%

Jennings play so far over the past three games makes him the slight favorite over Holhauzer. (I made this timestamped non-editable prediction in advance of tonight’s 14 Jan game as proof. Also note that the publication date of this post is before that of the game.)

Update 14 Jan 2020 @ 9:01PM: This resolved correct. I bet $45 on this  resolving incorrectly (in favor of my prediction) and lost the bet. …I guess I won’t end up like Bryan Caplan.

Other Miscellaneous Bets

Update 16 Dec at 1pm ET: New bet of $10 with David Manheim about which one of us will better predict on Vox’s Future Perfect questions.

Meta-Prediction: I win the above bet with David Manheim - 20%

Added 17 Jan 8am ET

He has a much better track record than me.

I put this on PredictionBook.

Could US Actions Topple Iran’s Government?: The four main options to establish regime change in Iran don’t work well: invasion (high reliability but high cost); air power (good to have no boots on the ground, but it requires near-flawless real-time intelligence to kill a head of state and then gives no control over who takes power next); economic coercion (crippling sanctions haven’t worked yet); and cooperation with local opposition groups (fail more often than they succeed generally… and not a viable option for Iran in particular).

Regime change rarely succeeds. When will the U.S. learn?: “Even after watching the chaos produced in Afghanistan, Iraq and Libya following regime change, some in Washington have continued to advocate similar policies toward Venezuela, Iran, North Korea and elsewhere. The belief that removing a foreign government can quickly and easily promote U.S. interests by force still resonates[. …] The problem, however, is that a resounding amount of research has shown that regime change rarely succeeds. Regardless of the goal, regime change mostly fails to produce better economic conditions, build lasting democracy or promote more stable relations to advance U.S. interests. From Haiti and the Dominican Republic in the 1910s, to South Vietnam in the 1960s, to Iraq in the 2000s, the United States failed to achieve these goals over 110 years of regime-change missions.”

At War with the Truth: The Afghanistan Papers: “A confidential trove of government documents obtained by The Washington Post reveals that senior U.S. officials failed to tell the truth about the war in Afghanistan throughout the 18-year campaign, making rosy pronouncements they knew to be false and hiding unmistakable evidence the war had become unwinnable.”

Getting Rich: from Zero to Hero in One Blog Post: “The standard line is that life is hard and expensive, so you should keep your nose to the grindstone, clip coupons, save hard for your kids’ college educations, then tuck any tiny slice of your salary that remains into a 401(k) plan. […] Mr. Money Mustache’s advice? Almost all of that is nonsense: Your current middle-class life is an Exploding Volcano of Wastefulness, and by learning to see the truth in this statement, you will easily be able to cut your expenses in half – leaving you saving half of your income.”

Killing your $1000 Grocery Bill: Shop at Aldi / Costco and don’t buy stupid stuff.

The Media Has A Probability Problem: The media keeps misinterpreting data — and then blaming the data. Handling uncertainty - including from errors of analysis and errors of interpretation - is important. The media is frequently overconfident. Journalists should be wary of ‘the narrative’ and more transparent about their provisional understanding of developing stories.

Eschew Sesquipedalian Obsfucation: “The best writers – like George Orwell – usually stick to short and simple words. […] Why doesn’t everyone follow Orwell’s rule? Harbaugh has a clean answer: If you’re a writer of moderate ability, you can’t make yourself look good using ordinary words. So you hide behind pompous language, demonstrating at least that you know more words than the average slob. In contrast, a great writer can sound brilliant in monosyllables – and those who can, do.”

Writing for On Balance: A Note from the Outgoing Editor: “Think about the key messages. What do you want your reader to come away with and remember from your blog? […] Use a compelling start. […] Get to the point quickly. […] Be concise. […] The ideal blog length is less than 1,000 words. No one can read much more than that with their coffee. […] Keep paragraphs short—no longer than 5-7 lines. […] Limit the citations. What do readers really need to know to be sure you have done your homework well, or where they can go for more information? […] Finish strong. Your concluding paragraph is important: what is the takeaway message for the reader? […] The reader is almost always right. After all, writing is about communication and you don’t really know what you’ve communicated until you get feedback.

Jeff Bezos Knows How to Run a Meeting. Here’s How He Does It: two-pizza team rule (don’t have meetings with more than six people), use a detailed memo instead of powerpoint, start the meeting with everyone reading the memo (otherwise they’ll just pretend to read it).

How to lose a monopoly: Microsoft, IBM and anti-trust: "A big rich company, a company that dominates the market for its product, and a company that dominates the broader tech industry are three quite different things. Market cap isn’t power. IBM ruled mainframes and Microsoft ruled PCs, and when those things were the centre of tech, that gave them dominance of the broader tech industry. When the focus of tech moved away from mainframes and then PCs, IBM and then Microsoft lost that dominance, but that didn’t mean they stopped being big companies. We just stopped being scared of them. For both IBM and Microsoft, market power in one generation of tech didn’t give them market power in the next, and anti-trust intervention didn’t have much to do with it. It doesn’t matter how big your castle is if the trade routes move somewhere else.”

The End of the Beginning: “[T]here may not be a significant paradigm shift on the horizon, nor the associated generational change that goes with it. And, to the extent there are evolutions, it really does seem like the incumbents have insurmountable advantages: the hyperscalers in the cloud are best placed to handle the torrent of data from the Internet of Things, while new I/O devices like augmented reality, wearables, or voice are natural extensions of the phone.”

Reality has a surprising amount of detail: “This means it’s really easy to get stuck. Stuck in your current way of seeing and thinking about things. Frames are made out of the details that seem important to you. The important details you haven’t noticed are invisible to you, and the details you have noticed seem completely obvious and you see right through them. This all makes makes it difficult to imagine how you could be missing something important. […] When you talk to someone who is smart but just seems so wrong, figure out what details seem important to them and why. In your work, notice how that meeting actually wouldn’t have accomplished much if Sarah hadn’t pointed out that one thing. As you learn, notice which details actually change how you think.”

Canceling: An emotional story of how cancel culture works by presumption of guilt (without any verificiation), abstraction (turning specific transgressions into generalized badness), essentialism (assuming a person is fundamentally bad because they made a transgression), pseudo-moralism, no forgiveness, transitivity (assuming people who support the transgressor are also transgressors), and dualism (assuming that all bad is bad and there are no shades of badness). Let’s be measured and nuanced.

Alcohol Is Killing More Americans Than Ever: The amount of deaths via alcohol is on par with death from opiods and death from guns and deserves a similar level of policy response.

Who’s behind those mysterious Colorado drones? Five theories.