There’s no point in listing my mundane views, so these are my unusual views. I’m more boring than this makes me sound.

Health

  • Doctors get too much credit for the doubling in longevity in the last 200 years: nutrition and public health measures (plumbing, water treatment, immunisation) explain more than individual medical therapies. (We know this because the largest improvements were before 1920, when medicine really began its stride.) This seems to still be true for the present day.
    NB: Morbidity might be different.
    (Confidence: 80%)

  • Massive amounts of medicine is based on bad or anecdotal evidence. RCTs very often find that age-old, popular, uncontroversial treatments have no average effect, or worse. Doctors often ignore the stats on treatments, relying on binary ideas of risk and contraindication.
    (Confidence: 95%)

  • But even the best evidence-based medicine is surprisingly unreliable, in the sense that most treatments won't work for most people. e.g. even morphine has a "number needed to treat" post-op pain of 2.9. i.e. on average, it reduces pain intensity by at least half for only every third person it is administered to.2
    (Confidence: 90%)

  • The UK does not add iodine to its salt; instead it is added to cattle feed and we blindly hope humans get some benefit if they drink a lot of milk. British vegans should take iodised salt, and perhaps also the omnivore population: several small studies from the 2010s found that levels in young women were about half the adequate level. Society depends on people like this. 1
    (Confidence: 80%)

  • In the general population, multivitamin use does not reduce mortality, and are even potentially harmful owing to overdosing beta-carotene and vitamin E. More generally, antioxidant supplements are not net positive. Many dietary supplements have no effect.
    (Confidence: 80%)

  • Reducing salt is a surprisingly bad treatment for high blood pressure. If you more than halve your sodium intake, the average effect on systolic blood pressure is a 1% drop (white people with normal BP), or up to 4.5% (white people with hypertension), and it has an array of bad effects on your hormones and lipids. (For reference, thiazide gives an 8% drop and ACE inhibitors give a ~7% drop.)
    (Confidence: 70%)

  • Vaping is a really remarkable public health measure: thousands of times less carcinogens, increased cessation change, massively reduced second-hand exposure, incredibly low cost. All of the reported acute deaths involve bootleg vaping fluid with excess vitamin E. Banning vaping and not smoking is a perfectly perverse policy which should be expected to shorten Californians' lives by thousands of years per calendar year.
    (Confidence: 70%)

  • Stannous fluoride toothpaste is better for your teeth than the common sodium fluoride; it's antimicrobial. (It used to taste a bit worse.)
    (Confidence: 80%)

  • "Prophylactic scaling" - where a dentist scrapes deep between your teeth despite no sign of gum disease, has little evidence of benefit, is unpleasant and expensive, and can damage enamel.
    (Confidence: 80%.)

  • Paracetamol (acetaminophen, Tylenol) is a bad choice for first-line pain relief. It's less effective than the alternatives: its number needed to treat is 3.5 [2.2, 13.3], worse than ibuprofen (1.7) or diclofenac (1.8), and worse than aspirin for migraine. It has an extremely low "therapeutic index": chronic liver damage can occur from 2 x 500mg tablets, and 8 tablets is dangerous. (Compare this to aspirin's 8 and 15 x 300mg.) About 20% of paracetamol-induced liver damage is accidental. NSAID alternatives cause problems for many people, but it's often possible to identify those at risk, and I see no reason to privilege avoiding gastrointestinal trouble over liver trouble.
    (Confidence: 90%.)

  • Smoke from fireplaces is an incredibly powerful risk factor for lung disease, much stronger per gas volume than cigarettes. The personal risk from a proper stove is limited, but neighbourhood effects can be severe.
    (Confidence: 80%)

  • The evidence for any health benefits from pre-exercise stretching is notably weak.
    (Confidence: 70%)

  • You should wear earplugs on the subway. Many of them run loud enough to cause permanent damage to your hearing: the London tube reaches 105 dB on some lines. This level of noise causes tinnitus reliably, and tinnitus may be associated with all kinds of mental health havoc.
    (Confidence: 80%)

  • Universal screening for rare things (breast cancer, depression, terrorism, whatever) is in general bad, because few diagnostic tests have good enough specificity to prevent this causing a horrific number of false positives.
    (Confidence: 80%)

  • Daylight Saving Time is an unacceptable public health burden. This is mostly down to the acute 6-16% increase in car accidents, but it seems to cause a 5% increase risk of heart attack and more for stroke.
    (Confidence: 70%)

  • Most people wash their hands in a way that doesn't help much. Needs to be >20 seconds, lots of soap, back and front, lots of friction. (Also, washing your hands properly a lot leads to dry skin and cracks, a major risk factor for infections...)
    (Confidence: 90%)

  • Around a third of people north of latitude 40 are deficient in vitamin D during winter. (e.g. UK) Big oral supplement doses work, though you may also need extra vitamin K2 to clear the induced calcium from your blood.
    (Confidence: 80%)

  • Zinc acetate lozenges are one of the few things shown to be effective against the common cold, reducing duration by maybe 24 hours.3 However, there are anecdotal reports of heavy use blunting your sense of smell.
    (Confidence: 70%)

  • Intermittent fasting (e.g. skipping one meal a day, but eating more in the other two) seems to be very good for you.
    (Confidence: 70%)

Science

  • I am extremely ignorant about myself. My preferences, my aliefs, my causes. So are you for you.
    (Confidence: )

  • It seems that reinforcement learning, game theory, evolutionary game theory, and theoretical market calculation are very close and in many places equivalent. The thing they all are special casing could be called "distributed optimisation". This is mind-bending, a perfect illustration of the absolute power and relevance of computational complexity, and maybe useful.

  • Academic publishing is not rational. Pre-publication peer review is weak; post-publication peer review is where it's at. http://blog.mrtz.org/2014/12/15/the-nips-experiment.html

  • It is wrong to believe on insufficient evidence. This might be because it reliably leads to moral harm, or might be in itself. It is very often wrong to not act on insufficient evidence.
    (Confidence: %)

  • Probability theory has normative force: epistemic norm.
    (Confidence: %)

  • Debate is probably bad. It encourages an undetectable form of irrationality, rewards aggression and wit over accuracy, and drives attention away from nonbinary, nondecisive evidence.
    (Confidence: 70%)

  • Ivy League online courses have a 95% dropout rate. You can learn from the greatest lecturers in the world, for free. Yet, when you withdraw the job-market signalling of a Degree, no one gives a fuck. Combined with the fact that most internet-connected people do not even enrol, we have grounds for a terrible fact: A huge majority of people are not intrinsically interested in learning.
    (Confidence: )

  • Freud was not a scientist. He did not discover the unconscious. He does not deserve most of the acclaim he received and still receives.
    (Confidence: 90%)

  • A great many of the most famous findings in psychology of the C20th and C21st are exaggerated or spurious. The replication crisis means that you should apply a sceptical prior, shrinking all nonpreregristered effect sizes by a factor of 2 to 10.
    (Confidence: 70%)

  • Juries should be replaced by judges, especially in cases subject to bias or complexity.
    (Confidence: 65%)

  • Many common arguments in defence of philosophy fail. Others don't: it defends you against bad philosophy, .
    (Confidence: 90%)

  • Having a degree in a topic means surprisingly little about your knowledge of it, your authority over it, your ability to apply it to appropriate domains. This is partly just due to forgetting.
    (Confidence: 90%)

Politics

    * Cultures are worth nothing except in how they improve the lives of their participants. (And to a much lesser extent, their spectators.) The same with language.
    (Confidence: %)

    * Non-retributive justice: Criminals are purely victims of circumstance (genetics, environment, personality. not choosing to make the choices they made). We should abandon punishment as a goal and instead focus only on preventing future harm.
    (Confidence: )

    * Most international flights have large externalities for both carbon and pandemic risk, and should be taxed accordingly.
    (Confidence: )

Grand

Technology is our only hope for some problems: individual death, species death, life after Sol. (Confidence: %)

The long-term future of humanity matters much more than anything else. (Confidence: %)

We need productivity growth, and probably output growth too. There isn't enough output in the world to support everyone yet. (Confidence: %)

Progress: Contrary to popular and elite opinion, the world has been getting better in key ways (poverty, violence, gender, disability, race discrimination, intellectual depth, freedom) for 70 years, and better in some key ways for 200 years. There’s a chance we could continue this to a dizzying degree. (Confidence: %)

Heuristics and biases: Humans are deluded in predictable and previously adaptive ways. Why we don’t make sense. Implies scepticism. (Confidence: %)

Scientific imperialism: Despite that, we sometimes succeed in knowing. It’s wrong to believe things on insufficient evidence. Technical skill is vital for successful thought and some kinds of action. Naturalism works methodologically and maybe ontologically too. (Confidence: %)

Effective altruism: outcome-oriented, maximizing, cause-impartial egalitarianism. You can’t reliably act morally if you don’t know the truth. (Confidence: %)

Longtermism: Most value lies in the future; the moral significance of our lives is dominated by our effect on that. Implies focussing on “existential risks”, things that could end the entire future at once. (Confidence: %)

Cosmopolitanism: The rich world’s relative inaction for the global poor is an enormous moral catastrophe. (Confidence: %)

Animal welfare: The suffering of nonhumans is also an enormous moral catastrophe. (Confidence: %)

Bioprogressivism: Nature is not amoral, above judgment. Nor is it obviously good.

Natural death is an enormous moral catastrophe _in itself_. (Ending it could lead to worse problems, like permanent autocracy, ecological collapse, but probably not unfixable ones.) (Confidence: %)

Consequentialism can capture the intuitions behind deontology and virtue ethics, but not vice versa. (Confidence: %)

Computers

  • Password managers defend you against several of the worst cybersecurity risks.
    (Confidence: 80%)

  • VPNs are highly imperfect and still worth £30 a year if you pick one of the battle-tested ones.
    (Confidence: %)

  • The answer to "What was the first computer?" is complicated. The usual answer, the ENIAC, is arbitrary.
    (Confidence: 90%)

Trivia

  • Most book reviews contain no critical thought, no more than the blurb plus several "Yay"s or "Boo"s. Trust a review in proportion to the amount of direct quotation in it.
    (Confidence: %)

  • Most vegan food is fine, it just doesn't have enough fat. Double the oil and become happy.
    (Confidence: 80%)

  • MSG is a relatively healthy and delicious ingredient for all kinds of cooking. (It contains 1/3 the sodium of table salt.)
    (Confidence: %)

  • This is the best method for preparing garlic on every axis.

  • Macs are severely underpowered for their price. Even factoring in longevity and usability, they are still a bad deal for people who like performance.
    (Thanks to Kaarel for pointing out this is no longer true)

  • The top decile of Japanese animation is better than the top decile of Western animation - visually, musically, and thematically. (It has worse average idiocy, perversion, and repetition, but it is easy to avoid these once you learn the symptoms.) This is true even though it is much less well-animated!
    (Confidence: 80%)

  • There is surprisingly little evidence that Turing committed suicide. There was no autopsy; nor was the supposed suicide weapon ever tested for poison; he had been using cyanide at home for electroplating; he was a messy person throughout his life, with unwashed hands.
    (Confidence: 60%)

  • Science fiction is important, since, unlike most fiction, it tries to direct our attention to the unprecedented, what we can change. It is also (contingently) less focussed on incompetence and wallowing. It is where we get to do a dry-run of future moral problems, which are the greatest moral problems.
    (Confidence: %)

  • Private insurance is not a scam - if you can't afford to replace what you're covering.
    (Confidence: 80%)

Speculative

  • PhD study might be powerfully harmful to the average student's mental health.

  • Talk therapy works a little. But in strange ways: the content, the theory of the practitioner, doesn't seem to matter. Maybe therapy is about 1) having a high-status clever person actually listen to you and 2) apply the outside view and explicit rationality to your problems. For many people, it is their only source of the animal lift (1), and explains why hierarchy is important.
    Testable prediction: larger effects from psychiatrists than counsellors, and from fancy office decor.
    (Confidence: 70%)

  • The campaigns against food waste and against plastic food packaging are laudable, but they have costs. Excess food production is a de facto emergency buffer against the many things that can cause supply chains to break down temporarily. Plastic food packaging is a major contributor to the secular decrease in food poisoning.
    (Confidence: 60%)

  • It might be a good idea to make governments pay minimum wage for the paperwork it assigns its citizens.
    (Confidence: 50%)

  • Very intense, very blue light could have good effects on people with seasonal depression, or even the cognition of people without it.
    (Confidence: 60%)

  • OPEC, the oil cartel, are one of the most effective environmental groups in history. By keeping the price of oil about x times higher than it would be under competitive pressure, they drove massive improvements in heating efficiency, and provided a de facto subsidy for non-oil energy development.
    (Confidence: )

  1. Using the median UIC in schoolgirls and the number of school-aged children with an insufficient iodine intake, the UK is now in the top ten iodine-deficient countries worldwide, positioned between Angola and Mozambique.
  2. However, NNT is a measure over a binary outcome, and may not mean what I take it to mean.
  3. The first review I cite is withdrawn, but its conclusions are the same as its main critic, citation #2.



See also