neumannOneProcess2020

Activity Space

Modularization enables pieces to be used in many different technologies

The tech tree is fractal

Technological roadmapping can give an idea of where technological evolution can go and how to get there

A heuretic’s impact is not endogenous to the technology but depends on the environment

The tech tree is a scaffolding inside the knowledge frontier

Technodiversity is important

The creation of a heuretic has two steps - conception and implementation

Top of mind

We don’t have good institutional ways to create multi-point changes

Phenomena are things that happen without a human consciously making them happen

Impact in innovation is analogous to fitness in evolution

The tech tree model of heuretics is a directed acyclic graph

Heuretic is a better word than innovation

The difference between specialists and generalists is fractal in nature

Science and technology advance through the same fundamental mechanism

Interfacing between closer neighbors on the tech tree is easier than between nodes that are farther apart

There is a higher probability of accidentally unlocking a closely related node on the tech tree

Untapped technical people near the end of their careers may have more potential than young people

Charlie Songhurst conversation 31 Aug 2021

Dispersion is a more appropriate word than dissemination or diffusion for heuretics

Uptimal solutions in science and technology are rarer than you think

The affordance of a heuretic dictates how it can connect with another node on the tech tree

Breakthrough fundamental science

The difference between specialists and generalists is not clean but depends on which nodes they can extend

Ring Species are situations where group A can mate with group B and B can mate with C but C cannot mate with A

The tech tree model suggests that more work on more root nodes is important

How do you incentivize people to work on nodes farther down the tree?

Each heuretic by default inherits the constraints, affordances, and assumptions of its parents

Observational technological phenomena

Adam Marblestone feedback on design

Science is discovering and exploring phenomena while technology is using a principle to exploit phenomena towards a use

 from airplane design to drug design

What are the different moves that can happen with a heuretic?

# §Tech Tree Model of Heuretics
Imagine a tree growing “up” with “root nodes” “lower” or “deeper” in the tree. The height is some monotonic function of time, but it definitely isn’t 1:1. 
- [[The tech tree is fractal]]
- [[The tech tree model of heuretics is a directed acyclic graph]]
	- [[Each node on the tech tree can recursively be decomposed into more nodes]]
-  [[The tech tree is a scaffolding inside the knowledge frontier]]
- [[Connections create new knowledge]]
- [[The tech tree creates new nodes either through idea sex or learning by doing]]
	- [[The upshot of learning by doing being a first class process is that the implementation is as important as where the idea came from]]
- [[Tech trees first showed up in the Civilization boardgame in 1980]] but [[There do not appear to be any intellectual connections between video game tech trees and history+philosophy of science+technology]]. [[Tech trees are intuitive]].

### Heuretics
- [[As a platonic ideal, research is an unscheduled quest for new knowledge and inventions with hard-to-predict outcomes]]
- [[Heuretics are either inventions or discoveries]]
- [[Heuretic is a better word than innovation]]
- [[Heuretics come out of people’s brains (TECHNOLOGY IS PEOPLE)]]
	- [[Heuretics cannot exist separate from humans*]]
	- [[New ideas need to come out of a single mind]]
	- [[Brains are wombs for heuretics]]
- [[The creation of a heuretic has two steps - conception and implementation]]
	- [[Conception is the imaginative design of a heuretic]]
	- [[Implementation is the actualization of the heuretic in the real world]]
- [[The process of creating a new heuretic node is like building something out of legos]]
- [[Science and technology advance through the same fundamental mechanism]]
- [[Modularization enables pieces to be used in many different technologies]]
- [[Modularization enables the higher level technology to improve continuously even if pieces improve discretely]]
-  [[The affordance of a heuretic dictates how it can connect with another node on the tech tree]]
-  [[Each heuretic by default inherits the constraints, affordances, and assumptions of its parents]]
-  [[Interfacing between closer neighbors on the tech tree is easier than between nodes that are farther apart]]
- [[There is a higher probability of accidentally unlocking a closely related node on the tech tree]]
- [[Uptimal solutions in science and technology are rarer than you think]]

### Dispersion
- [[Dispersion is the process of a heuretic impacting the world]] ::needs work::
- [[Dispersion does not happen instantly and often requires a lot of work]] ::needs work::
	- [[List of things that affect dispersion]]
- [[Constraints limit the impact and dispersion speed of a heuretic]] ::needs work::
- [[Dispersion is a more appropriate word than dissemination or diffusion for heuretics]]

### Specialists and Generalists
-  [[The difference between specialists and generalists is fractal in nature]]
- [[Modularizing ideas and specialization are the same thing]]
- [[The difference between specialists and generalists is not clean but depends on which nodes they can extend]]
- [[Fast moving fields have lots of unconnected nodes]]

### Other Phenomena
- [[General purpose technologies are technologies that are lower on the tree]]
- [[Macroinventions are the tail end of the impact power law distribution clashing with our intuition]]
- [[The phenomenon of seemingly trivial technologies becoming incredibly important happens because there is fog of war on the tech tree]]

## Tech Tree is analogous to evolutionary tree
- [[Impact in innovation is analogous to fitness in evolution]]
-  [[A heuretic’s impact is not endogenous to the technology but depends on the environment]]
- [[There are many heuretic environments]]
- [[Ring Species are situations where group A can mate with group B and B can mate with C but C cannot mate with A]]
- [[There are scientific and technological problems that make other technology advances irrelevant by altering the environment]]
- [[In the same way that species don’t suddenly replace each other, technologies and ideas don’t suddenly replace each other]]

## Upshots
-  [[Technodiversity is important]]
-  [[The tech tree model suggests that more work on more root nodes is important]]
- [[How do you incentivize people to work on nodes farther down the tree?]]
- [[People are incentivized to work on nodes higher up the tech tree]]
- [[Systems must take performance hits to get out of local optima]]
- [[The tech tree model suggests that explicitly drawing out tech trees is important]]
	- [[The structure of the tech tree can inform actions that are short term suboptimal but long term useful]]
	- [[Roadmapping could be thought of as backwards-chaining together a series of conception-only heuretics]]
- [[What does technological genetic engineering look like?]]
- [[What does technological husbandry look like?]]
- [[What does technological mutation breeding look like?]]

### References
- [[neumannOneProcess2020]]
-  [[Breakthrough fundamental science]]
- [[Stagnation and Scientific Incentives]]
- [[arthurNatureTechnologyWhat2009]]
- [[narayanamurtiCyclesInventionDiscovery2016]]
- [[Road-mapping Biology - Article]]
- [[Adam Marblestone Podcast 2020-03-31]]
- 

[Web URL for this note](http://notes.benjaminreinhardt.com/§Tech_Tree_Model_of_Heuretics)

[Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/§Tech_Tree_Model_of_Heuretics)






§Tech Tree Model of Heuretics

# Each node on the tech tree can recursively be decomposed into more nodes
At the end of the day, [[New ideas need to come out of a single mind]] so the recursion base case is when the heuretic isn’t viable outside of one person’s brain. This is why [[Brains are wombs for heuretics]]. 

The ‘viability condition’ is admittedly vague. Something can be inviable away from one person’s brain for different reasons. 

In one situation they’ve actually created something that only they can create. For example a glass blowing method that only one person can do. Does that count as a heuretic? In their mind they might literally be unable to separate out the piece that’s different from how everybody else does it, even though they obviously didn’t reinvent ever piece of glass blowing. This example suggests that heuretics can be much more complex and non-modular than we would like to think. [[There are many pieces of equipment or tacit knowledge that only exist  in one or two places in the world]]

What about the situation where someone has a possibly really powerful theory in their head but it’s extremely hard for people to verify or build off of it? This seems to be the case with [[Stephen Wolfram’s Theory of Everything]] and [[Eric Weinstein’s Theory of Everything]]. Where is the line between them and crazy people? At the end of the day, doesn’t a heuretic need to unblock other people in some way? 

Combining nodes can lead to properties that neither of the original nodes contained. [[Emergent behavior happens when a system has properties that none of its subsystems have]]. The emergent behavior is another hint that heuretics can be much more complex and non-modular than we would like to think. 

### Related
- [[Design for roadmapping system]]
- [[Roadmapping could be thought of as backwards-chaining together a series of conception-only heuretics]]
- [[Every time you zoom in on a fractal, it presents the same pattern on a different scale]]
- [[The tech tree is fractal]]
- [[The creation of a heuretic has two steps - conception and implementation]]
- In [[And Suddenly the Inventor Appeared]] it states that every technical system has a super system and sub systems.
- A core piece of [[arthurNatureTechnologyWhat2009]] is the systematic hierarchical nature of technology
-  [[Module assembly and evolution are two ends of a spectrum]]



[Web URL for this note](http://notes.benjaminreinhardt.com/Each_node_on_the_tech_tree_can_recursively_be_decomposed_into_more_nodes)

[Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/Each_node_on_the_tech_tree_can_recursively_be_decomposed_into_more_nodes)




Progress frontiers are geographically fractal

What are the properties of maps?

There are two levels of institution

Technological and scientific paradigms are each based around exploiting a specific phenomenon

Scientific inquiry is an abstracting process while engineering design is a narrowing process

Maps are multi-scale

olsonManagingModesEffort2020

How can you describe the difference between learning by doing and idea sex?

Each node on the tech tree can recursively be decomposed into more nodes

Conservatism and Progressivism both have good and bad sides

Identity categories are fractal

# Every time you zoom in on a fractal, it presents the same pattern on a different scale
Fractals are Multi- Scale patters
## Questions
## Related
* [[Progress frontiers are geographically fractal]]
## References


[Web URL for this note](http://notes.benjaminreinhardt.com/Every_time_you_zoom_in_on_a_fractal,_it_presents_the_same_pattern_on_a_different_scale)

[Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/Every_time_you_zoom_in_on_a_fractal,_it_presents_the_same_pattern_on_a_different_scale)




Specialists create knowledge within fields

Idea sex probably happens more often in generalists

Notes that get screwed up somewhere in the pipeline

# Generalists create connections between high-degree nodes 	
This note was originally “Generalists create connections between disciplines.” That statement is imprecise because [[The tech tree is fractal]] so the line between disciplines is mushy. Roughly, the higher the degree of the nodes that someone is connecting ([[Connections create new knowledge]]) the more of a generalist they are. 

### Related 
- The idea of elaboration in [[How to Take Smart Notes]]
- [[Connections create new knowledge]] 
- [[The difference between specialists and generalists is fractal in nature]]. 



[Web URL for this note](http://notes.benjaminreinhardt.com/Generalists_create_connections_between_high-degree_nodes)

[Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/Generalists_create_connections_between_high-degree_nodes)




§Program Design

Most roadmaps suck

Good Simulations could be the ‘why now’ of roadmapping tools

# Good roadmaps work backwards from a goal
Working backwards from a goal prevents ideas from being locked-in by how people think about the problem right now. 

In a way, thinking backwards from a goal is what people mean when they talk about first-principles thinking. The usual interpretation is to start with physics but I would assert that while tying back to physical laws is important to ‘first principles’ thinking, it’s actually the end. Or perhaps ending each step with checking “does this violate physics?” Is an important part of the process.  Working backwards from a goal is ‘first principles’ thinking in the Aristotelian sense of not working off of the work of others, which *is* critical to working backwards. 

The working backwards approach contrasts with the map-and-then-roadmap approach, where you would survey the entire landscape and *then* create a path through it. [[Roadmaps have two pieces - the map and the path]] . In reality, the approach is probably a deeply impure mixture of the two approaches - jumping back and forth between working backwards and describing the present state of the world. [[Understanding a thing has to do with intuiting the affordances of a thing]] and in order to understand the affordances of the area you’re playing in you need to at least engage with the way it does work now.  

In a way there is a tension between working forward and working backwards. If you purely work backwards you ignore the accumulated wisdom of one or more disciplines, which is considerable if those disciplines are valuable in the first place. [[We stand on the shoulders of too many giants to give them all credit]]. And to compound the problem, often that wisdom (especially wisdom on the [[Knowledge frontier]]) is not well modularized and composable, so it’s hard to know what’s important and what’s ignorable without knowing it. At the same time, starting with “the state of a field” can be a trap in a couple of ways.  [[The tech tree is fractal]] so you could actually spend an infinite amount of time digging into exactly what the frontier looks like. Additionally if you spend a lot of time immersed in current paradigms you’ll start to see the world through those paradigms, which is problematic for going after goals that almost by definition have not been reached by the people who work in those paradigms. 

This tension between working backwards and forwards is embodied any time someone approaches a problem outside of their ‘area of expertise.’ It is perhaps best illustrated by physics. Physicists are constantly bombarded by theories from quacks who think that they have found a new theory of <life, the universe, and/or everything.> At the same time, people who in the past *have* found new theories have on occasion come from non-traditional backgrounds and *were* regarded as quacks, until they weren’t. 

Perhaps the correct way to think about roadmapping is actually similar to a backpropagation algorithm, where you do many passes going both backwards and forwards. You

Not working backwards from a clear goal is one of the ways that [[Most roadmaps suck]].  A big reason they don’t do this is that working backwards is *really hard.* It forces you to consider important questions about something that doesn’t actually exist. Asking questions about “made up things” isn’t a mode of thinking we’re used to. 

[[The Architected Innov􏰂ation Institute - Engineering Scientific Revolutions]] asserts that working backwards from a goal helps you evaluate possible paths to that goal with an unbiased eye. 

[[Treating Science Fiction as case studies from a future history]] might be one way to make considering important questions about things that don’t exist. 

Thinking about hard-to-think-about questions regarding things that don’t exist seems like an actual place for tools for thought. [[You could build tools to make program design more efficacious]]. 

Another framing besides ‘goals’ is answers to the question [[What does ‘working’ mean?]] for this particular program?

One thought is that perhaps you [[Work backwards with questions and work forward with answers]]?

### Related
- [[A lot of research feels like it is building lego pieces without any idea how they’ll be used]]
- [[Roadmapping could be thought of as backwards-chaining together a series of conception-only heuretics]]
- [[Technological roadmapping can give an idea of where technological evolution can go and how to get there]]
- [[Luke Durant Email 18 Sep 2020]]



[Web URL for this note](http://notes.benjaminreinhardt.com/Good_roadmaps_work_backwards_from_a_goal)

[Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/Good_roadmaps_work_backwards_from_a_goal)




The tech tree creates new nodes either through idea sex or learning by doing

Learning by doing and idea sex happen through the same mechanism

# How can you describe the difference between learning by doing and idea sex?
# How can you describe the difference between learning by doing and idea sex
The difference between [[Learning by doing]] and [[Idea sex]] might have to do with *where* those ideas come from, but the more I think about it that seems like BS. Learning by doing has the sense that there is a lot of experience.

If [[Connections create new knowledge]], maybe the difference is *what* is being connected. Because at the end of the day it can’t be *how* they’re connected. [[Learning by doing and idea sex happen through the same mechanism]]

The distinction is tricky because you could argue that it is actually idea sex because the inventors didn’t just blindly try every shape. They probably went off of some intuition around symmetry, and maybe even used calculus to find the right shape.  Similarly, there are probably many pieces of knowledge that Henry Ford and colleagues brought to bear to get moving slaughterhouse technology to work with cars. 

The resolution of this has something to do with the fractal nature of the Tech tree and also the concept of modularity.  Some options for the difference:
	- Perhaps the difference is substitutability? In learning by doing, you create something that is strictly better than the thing before it. Learning by doing doesn’t change something’s function.
	- Perhaps the difference is modularity - if someone improved non-assembly-line car manufacturing or meat moving technology, that would make the moving assembly line better, but improving calculus wouldn’t make the shape of the rocket engine better. This is pretty weak. 
	- It might also be a qualitative thing where because it’s fractal you just look at whatever level you care about. 
See:
	- [[Modularization allows different groups to work on different pieces]]
	- [[Modularization enables pieces to be used in many different technologies]]
	- [[The tech tree is fractal]]
	- [[Every time you zoom in on a fractal, it presents the same pattern on a different scale]]

### Related
- [[The tech tree advances by idea sex and learning by doing]]
- [[Idea sex]]
- [[Learning by doing]]

[Web URL for this note](http://notes.benjaminreinhardt.com/How_can_you_describe_the_difference_between_learning_by_doing_and_idea_sex)

[Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/How_can_you_describe_the_difference_between_learning_by_doing_and_idea_sex)






You can’t cut off just one tail of a distribution

If new knowledge is combinations of other pieces of knowledge, can macro inventions in the mokyr sense exist?

# Macroinventions are the tail end of the impact power law distribution clashing with our intuition 
# Macroinventions are the tail end of the impact power law distribution clashing with our intuition 

In [[mokyrLeverRichesTechnological1992]], [[Joel Mokyr]] describes macro inventions as `a radical new idea without clear precedent emerges more or less ab nihilo` . There are many inventions that *feel* this way, but if you look closer there always is a clear precedent. [[Thoughts don’t spring fully formed out of heads]].  This apparent contradiction has led people to criticize macroinventions as a concept ([[Macroinvention vs. Microinvention?]]) and even Mokyr shifted his definition. 

But it really does feel like there is a difference between inventions like The Telegraph! The Power Loom! and inventions like the see-through umbrella. How do you resolve a situation where there *appears* to be two categories but inspection shows no clear line? [[Continuous changes can lead to discrete differences]]! Here, the continuous process is just a power-law distribution of invention impacts and the threshold is where it crosses the human intuitive lognormal distribution. 

The fact that you seem to be able to infinitely zoom in on macroinventions is one piece of evidence that [[The tech tree is fractal]]

![](../BearImages/5E70EA2E-C5F6-4742-B1D0-905E3152910A-68756-000639A9323E93AE/59800AFB-D4FB-4775-8E2C-D89B9252FBD8.png)

[Web URL for this note](http://notes.benjaminreinhardt.com/Macroinventions_are_the_tail_end_of_the_impact_power_law_distribution_clashing_with_our_intuition)

[Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/Macroinventions_are_the_tail_end_of_the_impact_power_law_distribution_clashing_with_our_intuition)






Legible games enable people outside of an institution to control it

People who have done a thing should be in charge of a thing

Creating legible interfaces is expensive

Fair games are self-limiting

Undiscovered Public Knowledge (Paper)

Maps help you figure out possibilities

Maps help you understand the state of something

A predictably successful idea is composed of known unknowns that that a team can execute against

Explicit rules are a substitution for learned pattern matching

Every simulation is an engineering approximation

Ontologies should be context dependent

‘Have your cake and eat it too’ life-games have led to a toxic society

Good quotes

# The Map is not the Territory



[Web URL for this note](http://notes.benjaminreinhardt.com/The_Map_is_not_the_Territory)

[Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/The_Map_is_not_the_Territory)




Just having the piece of literature doesn’t solve the problem

As a platonic ideal, research is an unscheduled quest for new knowledge and inventions with hard-to-predict outcomes

Academic research is often just the concept half of creating a heuretic

Technology transfer is a mystifying abstraction

# The creation of a heuretic has two steps - conception and implementation
1. [[Conception is the imaginative design of a heuretic]]
2. [[Implementation is the actualization of the heuretic in the real world]]

[[The process of creating a new heuretic node is like building something out of legos]].

More generally, the creation of a heuretic always has two steps. First you need to conceive that two other heuretics can be brought together and how they might connect - this is the eureka moment, the hunch, the idea, the concept. Second you need to actually connect the heuretics to create a new heuretic that can be used by someone else as a single unit, this is the implementation.  For now let’s call the first step “concept” and the second “implementation.”

Concept and implementation vary wildly based on the heuretic. In math, the concept is the idea that you could prove something and generally what the pieces of the proof would entail while the implementation is writing down the proof itself. In scientific-method-based science, the concept is the hypothesis and the implementation is actually doing the experiment and analyzing the results. In technology, the concept is the design and the implementation is actually building the thing. 

Similar to many other parts of the tech tree ([[The tech tree is fractal]]) lines between concept and implementation are fractal.  [[Every time you zoom in on a fractal, it presents the same pattern on a different scale]].  So while a design could be just the concept step of a heuretic, the design can also be a heuretic itself. In the case where the design is in itself a heuretic, the “hunch” of which pieces should go together in the design is the concept. Implementation is just recursive conception until you can get to a conceived action that someone can do without coming up with a new thought. ([[Each node on the tech tree can recursively be decomposed into more nodes]])

The amount of work needed by each steps varies significantly based on the heuretic. Some heuretics resemble [[Problems where it is easy to know whether an answer is correct but hard to get to that answer]] in that once you have the concept, implementation is trivial. For other heuretics, the majority of the work is in the implementation. To go back to the Lego analogy, it’s like knowing exactly where a piece needs to go but it’s way in the back of some small opening that requires you to either crack open the entire structure to use tweezers or something. The variable division of work between concept and implementation is one of the reasons that “ideas are cheap.”  In many cases the division of work between concept and implementation is variable. The more precise the concept is, the easier the implementation. [[Make precise predictions]], [[Precise communication enables other people to make decisions]], [[More precise descriptions of technology could enable faster progress]]. 

Just as you need to fiddle with a lego piece to get it to snap into place even when you know exactly where it needs to go, you often need to create new heuretics during the implementation of a heuretic. These new heuretics are often [[Learning by doing]]. 

Clean, legible interfaces to heuretics make both concept and implementation easy and bad interfaces make them hard. Legos are such a joy because they’re basically the paragon of clean interfaces. Legible interfaces make it easier to grok the affordances of a heuretic, which makes it easier to conceive what you can combine with it. So while  [[Creating legible interfaces is expensive]], if you want your heuretic to be used, it is important to focus on legibility. 

This notion that conception requires understanding of constituent heuretics ( [[Understanding a thing has to do with intuiting the affordances of a thing]] ) is important because it helps explain why people who have fuzzy ideas of how things work have trouble implementing their ideas.

### Related
- [[Connections create new knowledge]]
-  [[§Tech Tree Model of Heuretics]] 
- [[Modularization requires legible interfaces]]
- [[The constraint framework is powerful because it shines a light on precise ways to explore new ideas]]
- [[It is important to precisely define problems in an area]]
- [[Brains are wombs for heuretics]]
- [[Heuretics that never leave the conception stage can lead to other heuretics]]
- [[Dispersion is the process of a heuretic impacting the world]]
- [[Academic research is often just the concept half of creating a heuretic]]



[Web URL for this note](http://notes.benjaminreinhardt.com/The_creation_of_a_heuretic_has_two_steps_-_conception_and_implementation)

[Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/The_creation_of_a_heuretic_has_two_steps_-_conception_and_implementation)




# The difference between specialists and generalists is not clean but depends on which nodes they can extend
Because [[The tech tree is fractal]], the difference between specialists and generalists is not “the number of nodes they can extend.” Instead the difference depends on the fractal degree of the nodes they can extend. Those who we think of as generalists are the people who are able to extend multiple high level nodes. *but* in order to do this, they often need to work with specialists who are able to extend the nodes that are fractally contained within that node.

More nodes-within-nodes are legible to specialists, while more higher-fractal degree nodes are legible to generalists. [[In order to extend a heuretic it must be legible to you]].

This also has to do with modularity ( [[Modularization allows different groups to work on different pieces]] ) in that given any node, the generalists understand how its interfaces work, while the specialists know the interfaces at least one level down. 

### Related
- [[§Tech Tree Model of Heuretics]]
- [[§Specialists vs. Generalists]]



[Web URL for this note](http://notes.benjaminreinhardt.com/The_difference_between_specialists_and_generalists_is_not_clean_but_depends_on_which_nodes_they_can_extend)

[Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/The_difference_between_specialists_and_generalists_is_not_clean_but_depends_on_which_nodes_they_can_extend)




There are scientific and technological problems that make other technology advances irrelevant by altering the environment

boydenArchitectingDiscoveryModel2019

Macroinventions are the tail end of the impact power law distribution clashing with our intuition

Good roadmaps work backwards from a goal

Generalists create connections between high-degree nodes

We can cleanly attribute philosophical ideas to people but we cannot do the same with science or technology

# The tech tree is fractal
[[Every time you zoom in on a fractal, it presents the same pattern on a different scale]] - in this case, every time you zoom in on a piece of knowledge, there are multiple pieces of knowledge that gave rise to it. 

From a theoretical standpoint, it’s intuitive that this is basically how human brains work. Our brains smush together different pieces of knowledge to create new pieces of knowledge. ([[Thoughts don’t spring fully formed out of heads]].) 

From an empirical standpoint, a fractal tech tree fits the observed phenomena:
- Fractal tech trees explain the power law distribution of impact in [[neumannOneProcess2020]]. Fractals give rise to power laws in general (see [[Self-Organized Criticality: An Explanation of 1 over f Noise]]) 
- Fractal tech trees explain how some inventions *feel* different than others but upon closer inspection actually have precedent just as much as other inventions.  [[Macroinventions are the tail end of the impact power law distribution clashing with our intuition]].
- Fractal tech trees explain why you can infinitely pick apart any category of technology.  You can break apart guns into rifles, pistols etc. and trace all of their histories, down to exactly what every human did over the course of history.

### Related
- [[The Map is not the Territory]]
- [[§Tech Tree Model of Heuretics]]

[Web URL for this note](http://notes.benjaminreinhardt.com/The_tech_tree_is_fractal)

[Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/The_tech_tree_is_fractal)






General purpose technologies are technologies that are lower on the tree

Technical Planning Principles

A specific focus for PARPA

General purpose technologies are more likely than other technologies to generate positive second-order effects

# There are scientific and technological problems that make other technology advances irrelevant by altering the environment 
# There are scientific and technological problems that make other technology advances irrelevant by altering the environment 

In the evolutionary analogy, the technology changes the fitness/impact of the technology by changing the environmental part of the equation.  [[Impact in innovation is analogous to fitness in evolution]]

These are different than root nodes because root nodes are basically [[General purpose technologies]] that unlock 

The term ‘root node’ term first came from [[Demis Hassabis: The interview]] in the context of using AI to solve all the other problems.  Is the AI a precursor node or is it an environmental factor?  A proxy question question is whether computers that are able to simulate something 

In the tech tree model of innovation, nodes are not truly “root” nodes. And because [[The tech tree is fractal]] it’s pointless to ask what the root nodes were so we don’t have to do any philosophical navel gazing. 



[Web URL for this note](http://notes.benjaminreinhardt.com/There_are_scientific_and_technological_problems_that_make_other_technology_advances_irrelevant_by_altering_the_environment)

[Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/There_are_scientific_and_technological_problems_that_make_other_technology_advances_irrelevant_by_altering_the_environment)




Unexploited Leonardo Ideas

You can create a program design discipline that enables better research and development

Adam Marblestone Podcast 2020-03-31

The possible upsides of a healthy program design discipline seem like they could be huge

# boydenArchitectingDiscoveryModel2019
- The “moonshot” didn’t require new science - just a lot of engineering
- Contrast with trying to go to the moon in 1600
- One approach to complex systems is to understand building blocks
- Not enough technology for seeing and perturbing complex biological systems
- Clear where the interfaces between braincellls is 
- Innovation requires some serendipity
- But what if we could search through possibilities in a high throughput way
- Nucleic acid sequencing = biological, chemical, optical fields
- Discipline around building new tools
- Question - how do you pick the class of problem that the tool will solve 
- Think about tools that have multiple uses - ::[[Modularization enables pieces to be used in many different technologies]]:: 
- Think backwards from fundamental capability ::Something with constraints::
- How well can you envision or roadmap out possible solutions?
	- Imagine the opposite of what everyone is doing
	- Undserstand what is going on in each activity in an area and write constraints on each one
- Need a wide collaborative network of people 
- Journal that republishes old results [[Unexploited Leonardo Ideas]]
- Continuously split and tile possible solutions by a property ::[[The tech tree is fractal]]::
- Literally describe a tree - tiling trees
- Used by Fritz Zwicky
- How do you decide which path on the roadmap is best?
	- Principle of applied laziness
	- Technologies that work well in complex systems are simple
`1600. If someone were to pro- pose to land on the moon back then, would it have been possible for scientists to resist the temptation to just go for it, and instead to take a step back and accelerate the invention of calculus, work out the laws of mechanics, and begin experiments on aerodynamics? `
### Published
## Questions
## Thoughts
- [[The biggest problem with 
## Related
- [[Book Review - Surfing Uncertainty - SSC]]
## Authors
- [[Adam Marblestone]]
- [[Ed Boyden]]
## Link
~[https://www.cell.com/neuron/pdf/S0896-6273(19)30286-7.pdf](https://www.cell.com/neuron/pdf/S0896-6273(19)30286-7.pdf)~




[Web URL for this note](http://notes.benjaminreinhardt.com/boydenArchitectingDiscoveryModel2019)

[Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/boydenArchitectingDiscoveryModel2019)




Connections create new knowledge

Modularization requires legible interfaces

Research has many orthogonal and non orthogonal classification axes

Dispersion does not happen instantly and often requires a lot of work

narayanamurtiGenesisTechnoscientificRevolutions2021

Human capital theory of stagnation

Punctuated Equilibria and Technological Progress

Efficiency biases systems towards false negatives

Human intuition treats all distributions as lognormal

Outdated science theories and outdated technologies can both remain on the Knowledge frontier

baldwinModularityDesignComplex2006

Could you get S-Curves to pop out of a tech tree and human perception?

Power laws occur when a random process triggers cascades or kills exponential processes

Why is there so little slack in mission-focused research?

Ignoring incentives, scientific and technological knowledge will have more impact if it is more public

Discontinuous progress investigation

Gaurav Venkataraman conversation 29 Jul 2021

Evan Miyazono Conversation 20 Apr 2020

# neumannOneProcess2020
- Calls out the dichotomy between the incremental improvements and big ideas
- [[kuhnStructureScientificRevolutions1962]] says that science goes between ‘normal’ and ‘revolutionary’ phases. The revolutionary phases *differ qualitatively* from normal science. <- Descriptive theory
	- Normal science = puzzle solving - reasonable chance of success, not uncharted territory
	- Revolutionary = revision to entire precise
- Before [[Thomas Kuhn]] was [[Karl Popper]] saying science was problem solving. See problem, come up with theory, test theory deductively and empirically. If test failed, discard theory. <- Normative theory
- [[kuhnStructureScientificRevolutions1962]] says proponents of normal science resist revolutionary science. 
	- Science depends on shared understanding and changing that is harder than changing a single mind
		- ::[[Learning a new thing you want to both learn the jargon and explain it like you are five]]::
- [[Max Planck]] - `“An important scientific innovation rarely makes its way by gradually winning over and converting its opponents: it rarely happens that Saul becomes Paul. What does happen is that its opponents gradually die out, and that the growing generation is familiarized with the ideas from the beginning.”` Science advances one funeral at a time
- Stubborness is not irrational 
	- Precession of mercury contradicted newtonian mechanics but science didn’t discard newtonian mechanics even though it knew they were wrong
	- Newtonian physics was the best we had
- Kuhn argument that refutations gather until a revolution happens like sand piles or avalanches
- Sand piles and avalanches are power law distributed ::THIS IS THE CRUX::
- incremental tech change:radical tech change::sustaining change:disruptive change :: normal science : revolutionary science
- Example of glass factory improvements
- Improvements might follow [[Zipf’s Law]] 
- ![](BearImages/9B36C336-E9FB-45A9-9339-FE3DD9E904AE-68756-0005D85EEC6FB9DA/ln-v-pld.png)
- Power laws seem bimodal because we expect a gaussian so we’re surprised on each end
- Most technology has a tech tree
	- ::[[Do any technologies have zero precedent?]]::
![](BearImages/D7960F77-C2A5-4FEA-A30E-63105A0D8363-68756-0005DFA5A01836A9/Screen-Shot-2020-03-24-at-13.37.29.png)
- [[Modularization allows different groups to work on different pieces]]
- [[Modularization enables the higher level technology to improve continuously even if pieces improve discretely]]
- [[Modularization enables pieces to be used in many different technologies]]
- 
![](BearImages/426D9C67-D29F-43AC-B8BA-7A7B970E6FDC-68756-0005D8C87CFBE5EA/Screen-Shot-2020-03-25-at-12.24.14.png)
Technology impacts things further down the tree
- If technology is distributed like a tree above, the impact of random improvements at different levels of the tree will result in an impact curve that follows [[Zipf’s Law]]
- Impact in this case depends on how many technologies further down the tree a technological discovery affects
- `We just tend to see the high frequency of small changes and the large impact of the unusual big changes.`
- Issues
	- Tree is probably not a regular three
	- Uptake is not immediate
- Model assumes people work across an entire level ::I disagree here - I suspect it works more like [[Each future node in the tech tree model has some function that maps effort in with a success probability distribution over time]]. Although there could also be an effect where it’s easier to accidentally find something in closely-related things - [[There is a higher probability of accidentally unlocking a closely related node on the tech tree]]::
- Belief in two processes leads people to set up two separate processes
	- If you take a single process and break it into two parts, use only the first part, you by definition censor the second part
![](BearImages/7F00A136-BC7E-4471-9A90-E27E3313BAD4-68756-0005D9F0CF3C7141/bhattacharya-packalen-diagram.png)

- If you’re being judged on quantity of breakthroughs you are incentivized to move towards the leaves or to the right on the curve
- Even if judged on both, mean time to breakthrough nearer to roots is longer ::This is related to [[The idea that you need to make a lot of money in software before you can work on sweet sci fi shit is false]]::
- If people work only on the top j layers, the average impact of science is *cut in half*

### Published
April 2020
## Questions
- [[Does the tech tree being a directed acyclic graph change the Zipfs law math?]] - No
- How does uptake fit into the picture - this has to do with the environment 
## Thoughts
- How is science evolution different than punctuated evolution?
	- The mechanism of evolution is the same regardless of external forces
- Power law model of innovation impact 
- If this is the case, tech trees are really important
- The fractal nature of the process seems super important [[The tech tree is fractal]]
	- But then if it is a graph and not a tree, does the fractal quality hold? 
		- Yes - [[Self-Organized Criticality: An Explanation of 1 over f Noise]] doesn’t require the tree structure - just multiple dimensions
- [[There do not appear to be any intellectual connections between video game tech trees and history+philosophy of science+technology]]
- Let’s assert that [[Science and technology advance through the same fundamental mechanism]]  
- [[The phenomenon of seemingly trivial technologies becoming incredibly important happens because there is fog of war on the tech tree]]
- [[General purpose technologies are technologies that are lower on the tree]]
- The tech tree model of scitech explains macroinventions as simply inventions that are on the surprising end of impact size - [[Macroinventions are the tail end of the impact power law distribution clashing with our intuition]]
- [[In the same way that species don’t suddenly replace each other, technologies and ideas don’t suddenly replace each other]]
- [[The tech tree model suggests that more work on more root nodes is important]]
- [[The tech tree model suggests that explicitly drawing out tech trees is important]]
- [[Technodiversity is important]]
- [[Learning by doing does not require explicit out-of-system knowledge]]
- Avalanche dynamics and evolution?
- A pushback is that technology is adopted differently in different places - I think this can be resolved with
	- [[Progress frontiers are geographically fractal]]
	- [[Describing an effect in an inventive manner is important for separating invention and discovery]]
- [[Dispersion does not happen instantly and often requires a lot of work]]
## Related
- [[§Tech Tree Model of Heuretics]]
### Cites
- [[arthurNatureTechnologyWhat2009]]
- [[For and Against Method - Including Lakatos’s Lectures on Scientific Method and the Lakatos-Feyerabend Correspondence]]
- [[Theory and Reality: An Introduction to the Philosophy of Science (Science and Its Conceptual Foundations series)]]
- [[The Age of Entanglement - When Quantum Physics Was Reborn]]
- [[Stagnation, scientific incentives, and funding models - Roots of Progress]]
- [[Stagnation and Scientific Incentives]]

### Other
- [[Punctuated Equilibria and Technological Progress]]
- [[Ring Species are situations where group A can mate with group B and B can mate with C but C cannot mate with A]]

## Authors
- [[Jerry Neumann]]
## Link
[One Process | Reaction Wheel](http://reactionwheel.net/2020/04/one-process.html)

[Web URL for this note](http://notes.benjaminreinhardt.com/neumannOneProcess2020)

[Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/neumannOneProcess2020)






raoGUTSGrandUnified2018

Where do ideas come from?

People like to fund projects

# olsonManagingModesEffort2020
![](BearImages/D5D9E905-C612-4F5E-A6B0-2454A5757D39-76832-000626DD4800F73B/D204A86B-6791-4676-93CF-C60E4F1A6A3A.png)
![](BearImages/68A86E34-2464-438C-844D-6C26F3794959-76832-000626E368A04A26/6A893012-596F-41E4-8224-246B7FBB19BC.png)
- The key insight on top of [[raoGUTSGrandUnified2018]] is that the different modes are hierarchical.
	- My expansion on this is that the modes are actually fractal — when you’re in Play mode there are sub-quests, explorations, or projects. [[Every time you zoom in on a fractal, it presents the same pattern on a different scale]]. A quest has sub-play, exploration, or projects. Etc. Even projects have sub-quests, explorations, and play. Even if the path forward seems straightforward, you might need to figure out a creative solution based on a new context, etc. (This is why even if a prototype works, manufacturing the thing is hard [[Manufacturing is important and often ignored]])
		- [[The tech tree is fractal]] and 
- Different modes call for different management styles ([[Research management matters]])
	-  Lean — know how to manage the *method* — “mercenary lean-six-sigma stonecutter types” in other words craftsmen dominate here. 
		- My extension would be that this is where specialists matter [[Specialists create knowledge within fields]]
	- Fat — know how to strategically reduce scope to do various lean things — focus on *leading talented people* -- more [[Generalists create connections between high-degree nodes]]
	- Divergent — roles rather than tasks
	- Convergent — alignement and assigning tasks 
- The trick is identifying and arguing for which situation is which
	- Recommended questions
		- What level of abstraction is the objective?

#### Published
#### Authors
#### Link
### Related
- Expands directly on [[raoGUTSGrandUnified2018]]
	- Shifts “muddling through” to “quest” from [[obengAllChangeProject1996]] which I like
- [[The distinction between areas, programs, and projects]]
- Cites
	- [[obengAllChangeProject1996]]
	- [[bahcallLoonshotsHowNurture2019]]

[Web URL for this note](http://notes.benjaminreinhardt.com/olsonManagingModesEffort2020)

[Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/olsonManagingModesEffort2020)






Benʼs working notes

The tech tree is fractal

Related

Links to this note