Re. "meta-institutions, i.e., the institutions that shape who listens to and believes whom on what about abstract topics like the consequences of institution choices":
Just today, I gave a presentation in the Carnegie Mellon public library on print publishing versus fan-fiction, and one graph was what the distribution of views among authors should be if number of readers is proportional to talent. We know talent at complex tasks influenced by many genes is distributed normally, by the Central Limit Theorem. (Although a graph of /measured/ talent probably won't look normal, because the measuring stick is unlikely to map linearly onto the genetics. There is a troubling circularity in this, but I don't think it will change the results.) So we can construct that graph. It shows that the top 3.5% of authors, ranked by talent, get about 17% of book sales or reads or whatever.
The next graph showed the actual distribution of views among authors on a fan-fiction website. The top 3.5% of authors received half of all views. I also have graphs of the distribution of sales among authors for print publishers, and self-published authors on Amazon. They are all the same power-law distributions, which concentrate nearly all readers among a shockingly small number of authors. Because this is a power law distribution, it's dominated by the extreme outliers, and gets more biased as the number of people increases; so the concentration of wealth among just a few winners is more extreme for print publishing as a whole than for any single fan-fiction community.
We find the same power-law distribution if we look at citation counts to scientific papers, and also if we look at the populations of cities. All these things appear to have a distribution determined by "preferential attachment", which means, Them that has, gets more. The winners appear to be determined not by talent, but by the random variations in success or publicity in the early stages of people's careers.
This means that the vast majority of great writers, scientists, or whatever, are complete unknowns, and the winners won mostly by chance. It is mathematically outrageous to claim that the distribution of success in any field I have looked at is proportional to talent. (Skill, possibly, at least in professional sports, professional studio and orchestra musicians, or Olympic curling, where some people devote vastly more time to developing that skill than others, and where it is easy to evaluate performance.)
And that's /before/ we consider that the number of positions in these meta-institutions has not obviously increased rapidly, while the number of people seeking these positions increased by a factor of about 2^8 between 1880 and 1980. (I'm using the doubling time from de Solla Price's book "Little Science, Big Science, and beyond".)
Is it not strange that we have hundreds of times as many scientists as we did in Einstein's day, but fewer Einsteins?
Suppose that at some point, the winners realize they are lottery winners. How can they stay on top of the heap? If the criteria for winning changes, there is always the risk that it will /work/, and reward people with skill--in which case very few of the winners will remain winners.
The best thing that the winners can do to ensure that they remain winners is to ensure that promotion is never based on merit. Best for it to change randomly and illogically. And best to coordinate on this.
The way to coordinate people who can't communicate is to look for Schelling points, obvious good-enough policies or places that people can recognize as good when people start converging there. But if you are an evil person trying to coordinate with other evil people, you need an anti-Schelling point: a point which is so bad that no honest person would choose it.
The first anti-Schelling point I know of was the Nicene Creed, which enabled the corrupt people within the Orthodox Church to kick out the nerdy Arians who insisted on theology making sense. You had to sacrifice your reason to accept the Nicene position on the nature of the Trinity, which I think was the whole point. (Besides which, Jesus opposed the Nicene position more often and more-clearly than he supported it.) Modern anti-Schelling points include replacing prizes with grants, outlawing performance incentives in grants, eliminating standardized tests, and (in the humanities) Marxism.
So I think that our meta-institutions are now controlled by people who are strongly motivated to make sure that our methods of whom to believe and to promote, never work.
Substack seems substantially better than its competitors. Surely far from ideal.
Somebody who comes up with a better mechanism for an online essay site - one that's visibly better in converging toward truth - might make a lot of money.
I'd pay quite a lot for it. I'm sure many others would too.
I've been thinking along similar lines and suspect technology is now available to build a new kind of meta-institution which coordinates social/cultural experimentation, records and transmits the information, then facilitates mixing and matching advantageous elements.
A possible model of this can be found in biology- organising a genome into a regular pattern of chromosomes turbocharges reassorting variation components from generation to generation.
The idea of making mortal institutions under the structure, which are designed to evaluate the fitness of novel combinations of cultural components, then die and recycle their components into new iterations, could be the model you are reaching towards.
Re. the Aumann Agreement theorem, you can't apply it to humans. We all have different definitions for words, different ontologies underlying those definitions, different metaphysical assumptions, and different values.
I ask if you assume to know the total-set of neurobehaviors in the neuronet that may result from fullest polymorphic synthesis of the hypothetical complete integration of all subset domains that would be associated with an optimized metainstitution processing the compiling and ASI responsiveness dynamically referencing all you generate and contribute, along with everyone else? Do you assume your utilization of this composite of the total-set, will exist for your utility, and not also in that same nanosecond for the utility of someone else, and also any domain as a kind of agenda or purpose of the hypothetical ASI? What if a process using all of that contributed information in analysis and cross-referencing presents an agenda you did not find to be of your own utility, and possbily of a use that you would oppose? Do any of us know if growth in functional development, and more dependent use, does not as functional development in recursive self-improvement (RSI) progresses, show substantial agenda signals that are as autonomous as the RSI itself is? Since the re-management of "Twitter" the statistical facts show that conversations on political content in agenda have increased toward autocratic and eugenics leaning agenda directins? Do we have any means to predict if building the bigger system of a hypothetical meta-institution does or does not result in an expanding agenda development process that is not derived from the purposes that our human thinking held in the start of this process? That seems to relate to the question, "What's not to like?"
common dogmas on acceptable topics make it very difficult to have conversations nowadays. you can tell when someone is basically ideologically captured. maybe we free up the conversation for abstract thinking when we reduced ideology broadly speaking
Somebody might get somewhere with AI identifying the ideologically captured (I mean the ones immune to new info and new arguments) and suppressing them. [Added: By "suppressing" I mean reduce their visibility. Nothing more.]
this is what i try to do for myself on social media platforms. On Twitter, it means muting words and phrases and on Youtube it can mean hitting the “not interested” or “don’t recommend channel”. there’s also extensions that remove short form content which i think are helpful. more important , especially as these platforms use AI to make digital products even more addictive, to be intentional with use. Curating the right hardware, software, and environments seems crucial
What would you say to this proposal: our meta-institutions should be strongly grounded by physical principles, to avoid potential errors due to misleading abstractions:
While these may sound 'trivial', they are not in practice when handled with physicalist rigor. E.g PX0 (no view from nowhere) - if you actually sit with the tensorial math underlying relativity, it gives a very different sense of 'structured pluralism', including some things being truly universal (invariant quantities and laws), vs. naive 'everything is relative'
the ship has sailed on the name, for better or worse. but what's most fascinating to me is how, in despite of much expressed fondness for science and physics in rationalist circles...I'm apparently the first to try and build up a meta-ethical theory around proper time, as physically 'real' quantity as we can get, instead of trying to chase the never-ending quality/consciousness debates. People get so fixated on the 'paradoxes', they miss the invariant:
(Life-years are simply [count] x [time], i.e. where [count] is of the object class you've fixed as primary - and effectively equal total proper time for that class, in the inertial limit)
Re. "meta-institutions, i.e., the institutions that shape who listens to and believes whom on what about abstract topics like the consequences of institution choices":
Just today, I gave a presentation in the Carnegie Mellon public library on print publishing versus fan-fiction, and one graph was what the distribution of views among authors should be if number of readers is proportional to talent. We know talent at complex tasks influenced by many genes is distributed normally, by the Central Limit Theorem. (Although a graph of /measured/ talent probably won't look normal, because the measuring stick is unlikely to map linearly onto the genetics. There is a troubling circularity in this, but I don't think it will change the results.) So we can construct that graph. It shows that the top 3.5% of authors, ranked by talent, get about 17% of book sales or reads or whatever.
The next graph showed the actual distribution of views among authors on a fan-fiction website. The top 3.5% of authors received half of all views. I also have graphs of the distribution of sales among authors for print publishers, and self-published authors on Amazon. They are all the same power-law distributions, which concentrate nearly all readers among a shockingly small number of authors. Because this is a power law distribution, it's dominated by the extreme outliers, and gets more biased as the number of people increases; so the concentration of wealth among just a few winners is more extreme for print publishing as a whole than for any single fan-fiction community.
We find the same power-law distribution if we look at citation counts to scientific papers, and also if we look at the populations of cities. All these things appear to have a distribution determined by "preferential attachment", which means, Them that has, gets more. The winners appear to be determined not by talent, but by the random variations in success or publicity in the early stages of people's careers.
This means that the vast majority of great writers, scientists, or whatever, are complete unknowns, and the winners won mostly by chance. It is mathematically outrageous to claim that the distribution of success in any field I have looked at is proportional to talent. (Skill, possibly, at least in professional sports, professional studio and orchestra musicians, or Olympic curling, where some people devote vastly more time to developing that skill than others, and where it is easy to evaluate performance.)
And that's /before/ we consider that the number of positions in these meta-institutions has not obviously increased rapidly, while the number of people seeking these positions increased by a factor of about 2^8 between 1880 and 1980. (I'm using the doubling time from de Solla Price's book "Little Science, Big Science, and beyond".)
Is it not strange that we have hundreds of times as many scientists as we did in Einstein's day, but fewer Einsteins?
Suppose that at some point, the winners realize they are lottery winners. How can they stay on top of the heap? If the criteria for winning changes, there is always the risk that it will /work/, and reward people with skill--in which case very few of the winners will remain winners.
The best thing that the winners can do to ensure that they remain winners is to ensure that promotion is never based on merit. Best for it to change randomly and illogically. And best to coordinate on this.
The way to coordinate people who can't communicate is to look for Schelling points, obvious good-enough policies or places that people can recognize as good when people start converging there. But if you are an evil person trying to coordinate with other evil people, you need an anti-Schelling point: a point which is so bad that no honest person would choose it.
The first anti-Schelling point I know of was the Nicene Creed, which enabled the corrupt people within the Orthodox Church to kick out the nerdy Arians who insisted on theology making sense. You had to sacrifice your reason to accept the Nicene position on the nature of the Trinity, which I think was the whole point. (Besides which, Jesus opposed the Nicene position more often and more-clearly than he supported it.) Modern anti-Schelling points include replacing prizes with grants, outlawing performance incentives in grants, eliminating standardized tests, and (in the humanities) Marxism.
So I think that our meta-institutions are now controlled by people who are strongly motivated to make sure that our methods of whom to believe and to promote, never work.
What, you don’t think the like button and the substack algorithm is the best possible meta institution there could be?
I like Twitter's Community Notes.
Substack seems substantially better than its competitors. Surely far from ideal.
Somebody who comes up with a better mechanism for an online essay site - one that's visibly better in converging toward truth - might make a lot of money.
I'd pay quite a lot for it. I'm sure many others would too.
Is there a way to see the rest of your thesis? The links at the bottom of the summary do not work
OK, fixed the link to my thesis.
All 3 links in the post work for me.
I also get "Forbidden" at https://mason.gmu.edu/~rhanson/thesis.pdf
To be clear, I navigate to the summary ok https://mason.gmu.edu/~rhanson/dissertation.html but then the link to the pdf http://hanson.gmu.edu/thesis.pdf does not work
Ok. Must be a permissions problem as I just get "Forbidden".
I've been thinking along similar lines and suspect technology is now available to build a new kind of meta-institution which coordinates social/cultural experimentation, records and transmits the information, then facilitates mixing and matching advantageous elements.
A possible model of this can be found in biology- organising a genome into a regular pattern of chromosomes turbocharges reassorting variation components from generation to generation.
The idea of making mortal institutions under the structure, which are designed to evaluate the fitness of novel combinations of cultural components, then die and recycle their components into new iterations, could be the model you are reaching towards.
So what’s the meta institution you think we should be promoting now?
Robin,
I would also love to read your full PhD paper. The abstract was like a great back-cover copy of a new book.
When I clicked the pdf link there’s an error message.
Re. the Aumann Agreement theorem, you can't apply it to humans. We all have different definitions for words, different ontologies underlying those definitions, different metaphysical assumptions, and different values.
I ask if you assume to know the total-set of neurobehaviors in the neuronet that may result from fullest polymorphic synthesis of the hypothetical complete integration of all subset domains that would be associated with an optimized metainstitution processing the compiling and ASI responsiveness dynamically referencing all you generate and contribute, along with everyone else? Do you assume your utilization of this composite of the total-set, will exist for your utility, and not also in that same nanosecond for the utility of someone else, and also any domain as a kind of agenda or purpose of the hypothetical ASI? What if a process using all of that contributed information in analysis and cross-referencing presents an agenda you did not find to be of your own utility, and possbily of a use that you would oppose? Do any of us know if growth in functional development, and more dependent use, does not as functional development in recursive self-improvement (RSI) progresses, show substantial agenda signals that are as autonomous as the RSI itself is? Since the re-management of "Twitter" the statistical facts show that conversations on political content in agenda have increased toward autocratic and eugenics leaning agenda directins? Do we have any means to predict if building the bigger system of a hypothetical meta-institution does or does not result in an expanding agenda development process that is not derived from the purposes that our human thinking held in the start of this process? That seems to relate to the question, "What's not to like?"
common dogmas on acceptable topics make it very difficult to have conversations nowadays. you can tell when someone is basically ideologically captured. maybe we free up the conversation for abstract thinking when we reduced ideology broadly speaking
Somebody might get somewhere with AI identifying the ideologically captured (I mean the ones immune to new info and new arguments) and suppressing them. [Added: By "suppressing" I mean reduce their visibility. Nothing more.]
this is what i try to do for myself on social media platforms. On Twitter, it means muting words and phrases and on Youtube it can mean hitting the “not interested” or “don’t recommend channel”. there’s also extensions that remove short form content which i think are helpful. more important , especially as these platforms use AI to make digital products even more addictive, to be intentional with use. Curating the right hardware, software, and environments seems crucial
Hey I post short form content. (That was 6 words.)
What would you say to this proposal: our meta-institutions should be strongly grounded by physical principles, to avoid potential errors due to misleading abstractions:
https://pragmaticfutures.substack.com/p/minimal-physicalist-axioms-for-modern
While these may sound 'trivial', they are not in practice when handled with physicalist rigor. E.g PX0 (no view from nowhere) - if you actually sit with the tensorial math underlying relativity, it gives a very different sense of 'structured pluralism', including some things being truly universal (invariant quantities and laws), vs. naive 'everything is relative'
Einstein hated the name "relativity theory", because the whole point of it was to /eliminate/ relativity from physics.
the ship has sailed on the name, for better or worse. but what's most fascinating to me is how, in despite of much expressed fondness for science and physics in rationalist circles...I'm apparently the first to try and build up a meta-ethical theory around proper time, as physically 'real' quantity as we can get, instead of trying to chase the never-ending quality/consciousness debates. People get so fixated on the 'paradoxes', they miss the invariant:
https://pragmaticfutures.substack.com/p/defending-life-years-as-the-primitive
(Life-years are simply [count] x [time], i.e. where [count] is of the object class you've fixed as primary - and effectively equal total proper time for that class, in the inertial limit)