17-07-26 Contrarian Ideas I Have

A lot of lift comes from taking seriously and acting on beliefs that other people acknowledge but that aren’t socially proofed. These are the ideas that are actually different.

Anti-tribalism. One of the surest paths to truth.
Creativity can be understood and systematized.
Similarity is the core of all knowledge, information & learning systems.
Models of structure in information, vision of future for information theory.
The tragedy around the input paradigm in machine learning leading to failure in transfer learning and metalearning.
The systemetizability of everything.
The importance of representation learning for alignment.
Need for new tools and approaches in spaces that are doomed as a function of their structure (Biology, in general spaces where the generative process can’t be computationally replicated and the search for a comprehensive understanding of the space is compute infeasible)
Importance of volatility over expected value at the systems level
Status is a flexible fitness function
Reference dependence destroys all happiness and suffering.
Complex consequentialism - subsume deontology and virtue ethics with long term focus and heuristic aware consequentialism
Map-territory applied to mathematics, science. Our representation of mathematics as an artifice with historical overfitting.
All of human rights, morality is a map-territory confusion
Entrepreneurship as a safe space to funnel productive rebellious people in a capitalist economy, and a way to make them feel responsible for their failure if they're not competent, internalizing failings of society
Successful entrepreneurship is an insanely high bar. Capitalism, like democracy, does a great job of getting people to view problems with their world as their own inadequacy
Status (Dominance / Prestige) allows humans' fitness function to efficiently adapt to the environment.
Conflation is how you indoctrinate
The key question in political philosophy is actually technological... it's about mechanism design, and how to design structures that take advantage of complex systems. Designing capitalism through the market mechanism. Designing the scientific method. Taking advantage of selection and evolutionary mechanisms. Instead of a political philosophy that's low level (say, have the government set prices), something that's meta (a mechanism for setting prices). A mechanism whose performance is better over time, as information spreads and as knowledge grows.
By measuring relative merit (rather than growth, or objective merit), students who are relatively better get the emotional positive feedback and students who are relatively worse (regardless of their long term prospects!) are given emotional suffering and punishments for their efforts.
Agreeableness is over-focused on peace in the short term, and under focused on peace in the long term
All value structures imply judgment, and create a hierarchy where people who align with the values are better than those who aren’t aligned.
Tragedy where people become more superficial when communities are larger, because particular relationships are less valuable (lack of scarcity)
The collective project of collecting idea by idea until we've modeled all the spaces as doomed and flawed. This stems from a feeling that I'm hearing yet another idea, yet another interaction... There are exponentially many interactions of a wide body of ideas. And so ideas become common and uninteresting.
The only hope is to automate the process of idea creation. Or to substantially simplify problem solving, and so surpassing the need for ideas.
We don't have a language for probability. And this actually makes it impossible to have a huge number of important conversations. Too many conversations look like "object x is an instance of class y", followed by "there are examples of x that are not in y, and so x is not in y", followed by "most x's are in y", followed by "that doesn't mean x has to be in y, and so x is not in y".
If there's objective meaning (and there's probably not objective meaning), then there's some form of meaning outside of humanity. Which means that that meaning can be expressed more purely and powerfully and cleanly in the universe than the current broken human instantiation and grasping at that meaning.
Objective meaning makes it possible to sacrifice humans or humanity for some greater goal.
These transhumanists who are overly focused on longevity / immortality have missed much of the point... the goal is to transcend the important part of the human, mostly our minds.
There's also the "human" part of transhumanism. I'm more interested in transcendence than transhumanism. Transhumanism feels weirdly anthropocentric.
Progress is Memes / Culture acting recursively on itself. People are a function of their culture / genome, not agents. And so a culture produces a person’s behavior, which impacts the culture, and a body of selection mechanisms produce changes in human society.
Rights are the new religion - same aesthetic rationalized differently
There's this weird sense in which biological justifications for behavior are necessarily correct. Many will cry, 'determining what should be is not the same as determining what is'. But the reality is that the thing that decides what 'should be' is something that came out of what is. It's a part of it. And it can be understood through a biological lens as well.
Instead of deciding what should be without awareness of the mechanism that determines that in us, we should focus on the way that we come to decide what should be and understand it deeply. And then see how we can change it, get outside it, outside it in a way that lets us adjudicate between different versions of deciding on what should be
Mathematics and algorithm need to merge. Formal mathematics should be computable. The difference between Marr’s algorithmic level and the mathematical level is important. In some cases, mathematicians need to care about whether or not their formalism is computable. There are many operations in computation that don’t accord with basic mathematical notation, which mathematical symbols would benefit from.

Source: Original Google Doc