Thanks for the feedback, this prompted me to add an afterword to the piece about the dangers of optimising too hard.

To me there's a pretty important difference between 'it is systematically bad to do anything about inefficiency' and 'sometimes inefficiency is protecting something important'. I strongly agree with the second one, but strongly disagree with the first one. I think that, empirically, there often are free wins, such as making learning more efficient. And sometimes inefficiency and balance is important, but sometimes it's not, and the skill is telling the difference.

For example, I think leisure time is great, but this is because it's time to relax, take breaks and do what I enjoy. Not just because it's time that I spend not being productive. If I notice a bunch of time in my schedule not being productive, but also not enjoying myself, eg commuting to work, finding ways to swap that for more true leisure time is a good way to reduce wasted motion

You might enjoy a post I wrote on the importance of leaving Slack and spare capacity in your life: https://www.neelnanda.io/blog/38-slack - essentially on the importance of leaving some inefficiency, so you can be flexible and reactive, and seize new opportunities.

The ceiling in my house is, upstairs, roughly one meter above my head. Downstairs, I've got maybe two?

That is "wasted space". It takes more energy to heat, and uses more building materials. Efficiency dictates that I do not need that extra airspace above my noggin, and thus it should not exist.

But it feels better.

When I lived in Berlin, the doors and ceilings felt as if they were built for a race of giants -- much taller, and much wider, then they needed to be.

And it was wonderful. It took some getting used to, but I could feel the difference.

Japan is full of inefficiency. There are people, human beings in pressed shirts and white gloves, that manually clear trains for departure on crowded platforms. Even when things aren't crowded, there is almost always a person on every train platform in Tokyo.

And that feels better, too.

Yes, the automated systems -- which exist! -- probably do more for safety. But I really like knowing that I can scream out for help, and another human being has a job description that involves "hear me" and "send aid." Because I'm an engineer, and I know full well how often we miss edge cases.

And, yes, this is less efficient. And my ticket might be a little cheaper if we had robots doing all the work. But those people do an important job. I am glad they are there, doing that job. I really don't want to those jobs "efficiented" away by a bunch of management experts who are chauffeured about in cars with tinted windows, and never ride the trains anyway.

Games are inefficient. But they make us happy.

Love is inefficient. One could make he argument that it would be better for society if computers assigned relationships based genetic/personality/economic/whatever criteria. But I don't think any of us want that, either. Gattica was a dystopia.

I'm not sure how, but we need to capture the idea that human happiness requires that every system have, built-in, a certain amount of inefficiency.

But even when you can operationalize it, it doesn't account for the chaos. If every car was made exactly so to be able to carry an exact amount of weight (max. passengers * median baggage) you would be in trouble when you went out to get some bags of cement for example.

The obsession with efficiency can become very inefficient very quickly.

Uh, no? Maybe humans actually add in safety. Why do you think we have pilots in airplanes?

Humans are underrated.

I really don't want to those jobs "efficiented" away by a bunch of management experts who are chauffeured about in cars with tinted windows, and never ride the trains anyway.

I am sure if you listen to beancounters, you won't save money and actually drive the system to the ground and make people unhappy.

I'm not sure how, but we need to capture the idea that human happiness requires that every system have, built-in, a certain amount of inefficiency.

You exclude human happiness from the definition of efficiency.

Love is inefficient. One could make he argument that it would be better for society if computers assigned relationships based genetic/personality/economic/whatever criteria. But I don't think any of us want that, either. Gattica was a dystopia.

Why would anybody think this would be efficient or that it works at all?

For a number of reasons. First of all, to act as a backup, assistant, babysitter, and replacement for the autopilot in modern planes. Exactly what GP says: to cover up for missed edge cases and unimplemented features. Autopilots are great and can do much, but not everything (yet?)

Then, there's a thing called “culpability”. You can’t blame a machine but you can blame a person when an accident does happen. Accidents can be happily resolved as “pilot error” without having to call back the entire fleet of aircraft.

Also, human passengers usually feel much more safe and calm when they know that their potentially fatal transport is controlled by a fellow human, just like them, instead of some soulless machine that can murder you all without so much as a blink if it goes awry. Some time ago elevators were not that safe and had elevator attendants, now they are safe enough for passengers to be fine with a computer controlling them.

Biggest issue with autopilots is that, well, they're part of the machine, the human comes in very important when the machine is partially crippled - we've already seen what happens to automation when critical sensors fail in the 737 Max.

> You can’t blame a machine but you can blame a person when an accident does happen. Accidents can be happily resolved as “pilot error” without having to call back the entire fleet of aircraft.

That's not really the case, air accident investigations won't blame the pilots just because they're there. They'll happily point out design defects, or maintenance failures, or pilot error, if it was actually a contributing cause.

Of course, plane manufacturers love being able to blame the pilots when their autopilot engages Kill All Humans mode.

They really should move the switch for that feature a little farther away from the "Cabin Attendant Call" button.

I believe this is called a "Porsche".

Absolutely.

> I am sure if you listen to beancounters, you won't save money and actually drive the system to the ground and make people unhappy.

Yes and no.

Economic realities are a thing. Organizations that stop producing enough income to cover their expenses soon stop being organizations.

The problem isn't necessarily the counting of beans, but rather how beans distribute over time.

There's a balance to strike there, and both the cultural and regulatory environments in the US push very strongly towards "acquire beans today, regardless as to what this does to tomorrow's bean-acquisition capabilities".

So, yes, the Counters of Beans get a say. Their input matters. But it's not the only thing that needs to be factored.

>And it was wonderful. It took some getting used to, but I could feel the difference.

Fwiw I concur, this was my first impression of Berlin the first time I went there too, and it felt amazing. Compared to NYC or the SF Bay Area where space is more at a premium. Berlin has the ability to build outwards and use space generously, where the other two are horizontal space-constrained and can only build upwards (and SF chooses not to).

Oh, absolutely. I would struggle to think of a worse way to do things. :)

"Acceptable inefficiency" is a coordination problem, which are nominally solved via protected markets.

To provide an example: let's say that you want all battery factories to operate cradle-to-grave with zero pollution. This is way more expensive than just dumping manufacturing byproducts into the nearest river.

No one manufacturer is going to do this on their own, because it'll put them at a substantial disadvantage compared to their peers.

However, if the market imposes that cost as an entry requirement -- to sell batteries, you must be certified pollution-free -- now you've leveled the playing field. Every player has the same requirements.

This isn't a trivial thing, of course.

You can't allow the shortcut of "move manufacturing to a place where we can fill the rivers with lithium and nobody in power is going to care".

This also introduces two new problems: regulatory capture and corruption.

If it's all-in cheaper to bribe, blackmail, or infiltrate the regulators, then that'll happen. And regulation is often used as a weapon against smaller or newer players.

But I'd wager those are easier problems to solve.

As long as it doesn't give you an urge to invade Poland, it's fine!

Meanwhile we truck our goods around while trains would be much more efficient. Trucks are only cheap because we subsidise them through the roads they are using. Few people are aware that road wear increases to the fourth (!) power with axle weight. That means cars or bicycles are even ignored in the calculation because they do so little damage, while trucks are responsible for nearly all of it. But they don't pay for nearly all of it – we do.

In the meantime trains have to maintain their own rail system.

Similarly with electricity - long time ago it was possible to electrify remote communities but now pulling fiber there is such a problem that we view satellites as better alternative?

Indeed. Efficiency for what purpose and by what metric? We all know that once a metric is applied things not covered by the metric get squeezed out.

Someone on here used the phrase "self-Taylorism" to cover this kind of activity, which I think is very apt. Taylorism was the beginning of manufacturing productivity research, but we've moved on from there.

It's also reminding me of the failures of high modernism and Corbusier's "the house is a machine for living"; the result of which was houses that were neat but inhospitable and inevitably customised in "inefficient" ways by those that actually lived in them to make them livable.

Systems need some slack.

Humanity is not a Borg collective.

That's ergonomics in a nutshell. Keep your head balanced on top of your shoulders, not tilted up or down, your shoulders relaxed, wrists supported and your torso and spine straight.

:)

Have you got a link or reference for this that I can explore? I'd be interested in reading what he had to say here.