Science without Validation in a World without Meaning

This comment x 1000. (Some) Engineers think they have it all figured out. And they do...because someone figured it out all for them, and they get to work with the benefit of that knowledge.

This is a polemical statement, I'm closer to an engineer than a scientists and I really like having a solid model to work from. It took me a long time to understand how useful approximations are to solving hard problems. Because I didn't want to know the approximations, I wanted a perfect model.

(Oddly, I came to understand the power of approximations in terms of electronics. Electronics are filled with rules of thumb that help you figure out how your circuit works and help to elucidate the behaviors that are important to you. But it took me awhile to get over myself and use them. It was honestly a huge road block to progress in learning.)

My own take-aways: getting knowledge (de novo, ex nihilo) and applying knowledge are two very important things, but are separate skills. Ideally you'd like to have both.

Secondly science is difficult because you are trying to figure out true things in the face of glaring uncertainties.

"Medicine is a form of engineering in which the physician takes some action (drug, surgery, etc.) to alter the behavior of a physical system, the patient. Its underlying science is biology. Hence our ability to characterize medical knowledge depends on our ability to represent and validate biological knowledge."

This is really very wrong. The 'physical system' is conscious and suffering, which excludes Medicine from even the most vaporous definition of engineering. Real progress in medicine is much more akin to exploration (in the Christopher Columbus sense) than engineering.

Anyway, this is a typical engineer writing about biology. They always make the same mistakes:

1. Ignoring evolution. 2. Treating mechanistic models as first class citizens with disdain for data (eventhough the latter in the form of DNA is the sine qua non of all life) 3. Failing to understand that "The simplest complete model of an organism is the organism itself."[1] as von Neumann noted.

[1] https://onezero.medium.com/the-future-of-computing-is-analog...

>You've now left the realm of science and entered the realm of politics.

All badness is the realm of morality and politics, not science. Science can predict how many people will die, it can't weigh this up against other consequences. Science can't make moral judgements, only inform them.

>"Lockdown badness" is something that is the result of poor government policy and actually has solutions--it is not immutable scientific fact.

A big part of the lockdown badness is the economic damage. This is an immutable economic fact: stopping most people working will mean fewer things are produced, and more of the existing things and infrastructure will be consumed. This translates into worse standard of living and quality of life for people. Weighing this against the health damage from the virus is a matter of politics/ethics, not of science.

What we can see clearly is countries with no widespread forced closure of businesses, like Korea, Japan, Taiwan and Sweden, are having much better economic outcomes compared to countries with strong lockdowns like the US, France and Italy (expected yearly GDP growth of ~0%, vs ~-5% for the lockdown countries).

> "The simplest complete model of an organism is the organism itself."

But that is trivially true about everything, including golf balls. The interesting question isn't if the model is complete, but if it is sufficient (or complete enough) to make predictions.

And in the case of golf balls we can make quite good predictions with a very simple model, mostly because we are only interested in if it ends up in the hole or not. But a complete model of a golf ball would need to include the exact dynamics of the different layers and the turbulence around it, which we just can't do. Luckily, we don't need to.

So what is a sufficient model of a living organism? That of course also depends on what we want to know. I am quite certain that modern egg farmers have, for their purposes, quite sufficient models of hens. They know how fast the hens grow, the effects of different amounts and types of food, how much heat they produce, and of course how many and how big eggs they lay, etc..

And it is of course possible that we need to know the exact wave function for the complete system that is the cell to model it sufficiently enough to, say, synthesise new drugs. But I don't think so.

I think the system the author is looking for is probabilistic programming. It allows you to write programs/models with uncertain factors. Then at the end you can observe a few variables and it automatically computes the expectation and variance all the input variables. Check out psisolver.org for example.