Magic Scary Machines

But now, large-scale computer models are seen as generating data in themselves. No longer are models judged by how well they reproduce data from the real world-increasingly, models provide the data.

As if they were themselves a reality.

Michael Crichton, Caltech Michelin Lecture – January 17, 2003

In the 17 years since Crichton delivered this address (you should really read it) we have moved past relying on large-scale computer models to keep us properly terrified. Now, any old spread sheet model will do:

“Terrifying” I suppose so, since basic math is a terrifying thing to about 90% of our conventionally educated population.

But that’s not what she means.

We have here a computer model, although the math you should be able to do in your head. The default settings are insane, of course. Most egregiously, using the CFR – case fatality rate – of 1.5% when talking about *infections* for a disease where 99% show mild or no symptoms, is at least incompetent, if not out and out fraudulent. IFR – infection fatality rate – is estimated by the CDC between 0.1% and 0.3%. Using those numbers (which embody another family of logical errors, but we’ll get to that in a moment) we end up with something like 195,000 to 585,000 deaths. Hey, that’s nothing to laugh at, but, way, way less than the idiotic 2.95M being proposed here.

The other obvious error – obvious to those of us familiar with data, data collection, and models – is the assumed homogeneity of the population. Turns out that the majority of the population, those under 50 and in reasonably good health, have an IFR of about 0.0001% – one in a million chance. (1) With this bit of information in hand, we should rejoice, throw away the damned masks, and throw a huge, sweaty, 2 week-long street party – for all those who are under 50 and in reasonably good health AND anyone else who would rather party and face a minuscule risk of death than cower like rabbits.

Because – pay attention here – after than 2-week party, we’ll have reached herd immunity! Huzzah! And it would cost us fewer deaths, maybe around 100, from COVID, far fewer than are going to kill themselves coming and going to the party, drinking too much and falling on their face, or in any of the myriad other mundane yet fatal ways healthy people routinely die.

Herd immunity is an interesting concept, the point where R – the number of people, on average, one infected person infects – falls below 1. At that point, the ‘epidemic’ crashes. COVID seems to have about a 2 week cycle, mostly, from when you get exposed to when you are no longer infectious. Some subset of that time is when you are actually infectious. I don’t suppose it’s a yes/no issue, either: I might imagine a person with a hacking cough and a constant sneeze for 2 days is more infectious in the sense of more likely to infect people, than that same person the 2 days before and after during which he is spared the cough and sneeze.

But, hey, I’m a data guy, so I want to know how it works before dismissing concerns as unimportant, which is what rolling all the individual data points up into one tidy number does. Perhaps an asymptomatic person who washes his hands is effectively not a vector, while even the coughing, sneezing person above is only one for 2 days – who knows? But making a model necessarily – necessarily! – requires repeatedly rolling up a whole pile of uncertainties and lacunas into one number, which seems unassailable and engenders far, far more confidence than it should.

Anyway: herd immunity may very well have long been reached. Overall deaths fell back to long-term levels weeks ago. The sane, science and history based projection from Day 1: this thing would peak within a couple months of starting, then die out with warmer weather and growing numbers of immune people. The lockup has simply prolonged the collapse stage, if it indeed did anything at all besides wipe out jobs and savings.

Give Crichton the closing statement:

The establishment heaps praise on false prophets and hurls mud at the real prophets, those who could change the world for the better cling to old certainties and prejudices and hide their ignorance by attacking others.

  1. This is also, by the way, the reason masks for the general population are ridiculous even if they work: reducing a risk by 50%, lets’s say, from in this case 0.0001 to 0.00005 is simply not meaningful; to reduce the spread and, theoretically, delay herd immunity among the excruciatingly low-risk majority while the tiny higher risk minority is already sequestered, largely in nursing homes, is insane.

Author: Joseph Moore

Enough with the smarty-pants Dante quote. Just some opinionated blogger dude.

2 thoughts on “Magic Scary Machines”

  1. Joseph

    Have you read Nicholas Nassim Taleb’s tweets on the pandemic? It’s too technical for me but I think you’ll find them of interest.

    So TL;DR computer models are only as good as the underlying coded assumptions. And even then they can’t take into account the human factor.
    So how then are they helpful?


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s