The Golden Horizon and Bronze Means
Against the Golden Mean
Diffusing Mean and Opposite
In an engineering ethics class in college, my professor often raised the concept of the Golden Mean. He would typically use the same example: “confidence is the golden mean between timidity and hardheadedness.” In the same class, we would discuss historical or hypothetical engineering case studies. Should a firm invest in a safety measure and eat the cost of downtime? Should a firm deliver products that are only half-baked to keep up with competition? These concepts may appear similar, but the conflation of practical optimization and a philosophical pursuit of ideal ethics could hardly be more confused.
Ethical dilemmas aside, all engineers know that their trade is a study in compromise. Engineering effort is directed towards trade studies or simply “trades” to select optimal strategies before beginning to execute any strategy.1
Engineering optimization is totally different from the pursuit of the golden mean of some moral value. An engineering trade for performance or an actuarial study of risk, whether they pose a hazard to human life or not, is ultimately a utilitarian objective! The golden mean seems to allow this inherently pragmatic analysis to contaminate the essence of virtue ethics— the goal of describing values that stand as guides regardless of the situation!
You might imagine a local optimum of kindness where marginal kindness leads to diminishing returns in philanthropy due to generosity towards unfair players who take advantage of the kindness and waste resources. However, there is not an optimal moderation of kindness that is most virtuous. This waste is as much a symptom of lack of wisdom, a lack of resources, or a lack of any number of other goods, as much as it is an "excess" of kindness.2
I put excess in scare quotes and not lack, because I think wisdom and kindness are both unilateral goods! In other words, the "excess" of a good is only so because of the circumstance. There are multiple ways to solve the problem. The pragmatic optimization is the addition of cynicism, but the optimistic optimization is to be more kind and more wise.3
Pursuing this behavior is following the God-like path. Though it's difficult for mortals to be able to do the God-like actions, I think it may be more tangible than it sounds at first. Orthogonalism means avoiding the illusion of trades as opposites.
Logic of Extrema
Despite all this, there is a strong logical argument for the unavoidability of tradeoffs. This comes from the question of disjoint optima of distinct criteria. This is explained very well in this conversation between Daniel Schmachtenberger and Bret Weinstein on the fastest, bluest car. Bret explains that the distinct extremes of blueness and fastness are almost certainly impossible to maximize in a single car. This is closely related to the fact that two random high-dimensional vectors or, similarly, two random words are nearly orthogonal. If marginal blueness sacrifices the most minute degree of speed, then it is impossible to make a car that is the simultaneously the fastest and the bluest.
Simultaneous optimization is possible only when the optima are exactly coincident. While most characteristics are independent, some are components of others, like weighing less and being less massive or being circumstantially and generally true. I would offer that God as the extreme degree of love is the coincident optimum of many other virtues that appear to be tradeoffs to us. Unlike the exclusivity of blueness and speed, the tradeoff between generosity and wisdom is an artifact of our constrained capacity.
Gold and Bronze
Characteristics offered as the extremes that generate a golden mean are often orthogonal axes of a trade where you must choose a bronze mean.
Even with the reality of tradeoffs, never lose sight of the ability to go for more instead. Never lose sight of the golden horizon.4
ORTHOGONALIST
I believe an extremely understudied question is what the limit of abstraction that can be pursued is.
Engineers can make a thing → trade ways to make a thing → write a philosophy on how to make engineering decisions → philosophize on engineering philosophy → …
Engineers can make a thing from a material → trade which material to use → search for new materials → study the search for new materials → …
Engineers can make to do something → make a tool to make it better → make tools that make better tools → make (tool-making)^2 tools → …
It’s not clear how to make progress on a (tool-making)^3 tool. Some would confidently say “math is a generator of an arbitrary number of levels of technology”. Those some are mathematicians and I don’t disagree with their characterization. However, this same description applies to the first blacksmiths whose production of iron gradually served as a generator of the next generation of materials. Second, the description of math as preceding an arbitrary number of levels of development doesn’t keep the mathematician from being consumed as an argument of the function of the next order! What is the production function of mathematicians? Automated proof assistants and recent AI projects are perhaps the first non-human creators of mathematical knowledge. Again, these can’t escape the question of “why not go out another level?”
I’ve been writing notes on this problem, which I call Chain Generality, for years and so far the best insight I’ve had is that an ancient Chinese proverb from Guan Zhong holds the right wisdom:
The plan for one year is to plant grain;
the plan for ten years is to plant trees;
the plan for a lifetime is to cultivate people.
There are conflicting sources online on who actually wrote this. See my thread with ChatGPT for sources from outside the internet’s anglosphere.
Looking at the seven deadly sins and (lesser known) seven virtues, I found moderation or temperance listed as a virtue. I don’t believe this refutes my arguments as most clearly demonstrated by the opposite of temperance listed as the sin of gluttony. Gluttony is immoderation, but is most directly a submission to impulse, a lack of control. I believe the “yes and” of using more of all virtues is clearly in line with the notion of control, even though God is the only case that can utilize the extremes of virtue simultaneously.
I’d also like to highlight the difficulty of opposites as I’ve explored previously:
Haste is a perfect example of an antonym hashing collision between the orthogonal virtue of patience and sin of sloth. You can calmly believe in yourself and trust that success will come without allowing yourself a day of sloth. As in the previous example of “excess” generosity, incorrectly interpreting a lack of intensity as patience rather than an excess of sloth can be lethal to a project. Complementarily, seeing any slowdown as sloth can lead to misplaced criticism.
Wisdom and cynicism may be two of the hardest characteristics to untangle.
The evolutionary and economic lenses are almost identical. Both also depend on a huge difference in the element of game theory.
Tradeoffs line on a budget curve. Going for more of more is going to a higher indifference curve. Reaching higher indifference curves on the same budget constraint is technological or evolutionary advancement.
Optimizing in an adversarial environment is much harder than in a static one. This is my one liner for why landing F9 is easier than self-driving. Even though every mechanical problem is harder for getting to and from orbit than driving down the road, dealing with other actors is incredibly challenging. Evolutionary conflict is yet another step harder since their objective is almost purely selfish, rather than marginally selfish.
In the technical domain, a horizon can be defined as the fundamental (first-principles) limit. The hardest engineering problems are those far from fundamental limits. Operating in this regime means that all thinking is trade space. I believe this is the most intellectually interesting part of my design work at SpaceX, one reason that nature is endlessly astonishing, and part of why the quest for AI is so exciting.