I think you dismiss the non-lockdown options too glibly. Like many, you seem trapped in the dilemma "lockdown and kill the economy vs no lockdown and kill the people". You, and many others, need to have a close look at what countries like Taiwan (and others) have done - no lockdown and few deaths.
Possibly there is an element of western arrogance in the way many western countries refuse to learn from Asian countries that have been far more successful in keeping their economies open and deaths low. Two key elements that are surprisingly effective: ubiquitous use of face masks, and the use of cheap, simple, rapid metrics to exclude people from schools, shops, workplaces, and public transportation.
Simple metrics like checking for cold/flu symptoms and taking temperatures are not infallible ways to determine if someone has the SARS-CoV2 virus. But they are very useful, and can be done at a rate that dwarfs the possible rate of PCR tests. As a result you can actually gather more information across the whole society, and the information is more current.
Lockdown is also ruinously expensive and unsustainable. In my simulations maybe $20M/life saved, $300K/capita over a year.
Remember we need only get R0(eff)< 1. We do not need to get it to zero. That means a bit more than a 50% reduction, maybe 60% from the base rate ~2.5.
The level of infection at which "herd immunity" occurs is IMHO very much up in the air. The simple SIR models tend to overstate the level needed, because they generally don't take into account the fact that in epidemics the super-spreaders get taken out, one way or another, quite early, and the remaining more cautious and safer people have a much lower effective R0. IMHO it may be as low as 15% or as high as 50%, but unlikely to be as high as 65%. Epidemiologists talk about epidemics dying out for mysterious reasons, and it is speculated it is something about the microstructure of society, like superspreaders and more generally the variance of infectivity/susceptibility of people and groups, that is responsible.
I think we have our answer to the Fermi paradox in our hopeless response to the CV pandemic. The median European country has had deaths/million more than 10 times worse than best practice (Taiwan etc). https://www.worldometers.info/coronavirus/#countries
Civilizations will arise when the species concerned is only barely able to manage the job. I think world history suggests that this is very true of us. The chances of being up to handling the much more complex, difficult challenges of going to the next level seem low.
You seem to have the idea that a programming language should define a certain set of abstractions and that is that. But to many one of the key powers that programming brings is the ability to define and model new abstractions. In addition to your list I also would therefore also require
If deliberately infecting yourself, consider taking measures to check and to optimize your immune response.
Nutritionally rich diet, protein, vitamin D3/sunshine, etc.
The phenotypes affect the environment e.g. via competition for resources. At the same time, prey species are evolving. It looks to me like this model works in a limit where the environment is large and/or in the short term, but breaks down beyond that.
It is an interesting connection. By the way another way you can look at evolution is that the organisms absorb information from the environment.
Could you expand on which momentum anomaly you tested? One type is cross-sectional momentum (buy the top 1/10th of stocks that went up and short the bottom 1/10th), which is subject to major periods of drastic underperformance. This is all well documented in the literature. The other type is using momentum on the market as a whole, perhaps switching between asset classes based on momentum. I would not think that you could assess a strategy based on 6 months of performance.
My view as an investor since the 1980s is that the EMH is true to a 0th approximation. However massive agency issues in the fund management industry leave room to outperform on a risk adjusted basis if your incentives are different from the average fund manager. Some anomalies are just not exploitable by fund managers/agents because they would lose all their funds under management after periods > 12 months of underperformance.
LW readers interested in the topic may like reading the Alphaarchitect blog. https://alphaarchitect.com/blog/
1. Most people tend to be right about 60% of the time when they feel fairly certain. If we apply this logic to your assumptions, the chance that they are all correct is approximately 0. Many things we were told about CV2 turned out to be wrong. That is a bit simplistic, but your analysis should take into account that your assumptions may not all be correct, and the consequences of this. For example what if young people have silent organ damage, as has been reported? What if immunity is limited, uncertain, or short-lived, as if often the case with corona viruses? Such errors could be very costly. In general the strategy of "pick the most likely scenario and bet erh farm on that one scenario" is a poor strategy.
2. By getting infected now, you are giving away much by way of option value. The value of getting immunized later, of having better treatment later, of having better and less costly methods of limiting infection, etc.
3. You are falling for the false dichotomy of lockdown versus uncontrolled pandemic. I suggest you have a close look at Taiwan, which has had approximately 1/700th the death rate of the US for example, and which did not have a lockdown. While Taiwan did make a fast start, Australia got down into the Taiwan range of active cases within about 5 weeks, and other countries could also do this with a brief lockdown.
Techniques used by Taiwan include contact tracing, strict controls on entrants to the country, enforcement of quarantine of cases, use of soft metrics like temperature and cold/fever symptoms with exclusion from schools/work/transport for the symptomatic, and others. They have selectively closed some high risk businesses like "hostess bars".
This problem of becoming fixated on one aspect of a problem or one one thing generally, "Einstellung" in German, is an important cognitive bias that is not talked about often enough IMHO. A common example these days is the notion that the USA has one problem, Donald Trump, and with him gone all would be right with the world again.
In my program I assume
Fraction of people compliant with the masks policy(asymmetric-distancing-fraction-compliant-fv) 0.7Fraction of infection that still gets through from mask wearer to other person (asymmetric-distancing-outbound-ineffectiveness) 0.3Fraction of infection that still gets through from non-mask-wearer to mask-wearing person (asymmetric-distancing-inbound-ineffectiveness) 0.8
Given this, and numerous other assumptions including no other measures taken, the death rate falls from 0.65% to 0.48% of the population. This is a good benefit but not a total solution.
If you have better numbers for mask effectiveness than the ones I guesstimated above please let me know.
The other main dubious assumption in my model (other than no other measures taken) is uniformity of people. I am adding some options on that tomorrow.
In the series of posts including this one http://www.overcomingbias.com/2020/03/expose-the-young.html Robin Hanson has explored this option, along with various scenarios of deliberately infecting young volunteers with a reduced dose (variolation). In other viral illnesses such as smallpox infecting with reduced doses can greatly reduce the severity of the illness and mortality.
One of the issues is the feasibility of isolating high risk people. How do you isolate people who need care (e.g. people in care homes / nursing homes) from young people, when young people are going to be looking after them? Another is if you want to get to herd immunity, you may not have enough young healthy adults for this (perhaps 60% are needed, while young healthy adults are about 40-45% of the community).
Good summary and exposition of the situation.
In "On Intelligence", there is a hand-wavy argument that actions as well as perception can be seen as a sort of prediction. When I read this, it kind of made sense. But afterwards, when I was thinking how I would implement this insight in code, I began to feel a bit unclear about exactly how this would work. I have never seen a clear exposition from Jeff Hawkins or Numenta (his company). This is surprising because I generally find he provides very clear explanations for his ideas.
Does anyone have a clear idea of how actions would work as some kind of prediction? A pointer to something on this would be good also.