It's an interesting data set. I'm sure a compression contest will discover a lot of correlations that we weren't expecting. But compression does not does not depend on the order of the data. If I find a correlation between crime and poverty, I can use either value to predict the other and get the same compression ratio. It does not matter whether crime causes poverty or poverty causes crime. Compression doesn't tell us.
A few years ago I researched homicide rates and gun ownership rates by country and was surprised to find a weak but negative correlation. But that doesn't tell us why. Does arming everyone deter crime, or does crime result in stricter gun laws? The data doesn't say. You can use either data point to predict the other and get the same compression. On Sun, Aug 6, 2023, 7:07 PM James Bowery <[email protected]> wrote: > > > On Sun, Aug 6, 2023 at 2:51 PM Matt Mahoney <[email protected]> > wrote: > >> On Sun, Aug 6, 2023 at 11:28 AM James Bowery <[email protected]> wrote: >> > On Sun, Aug 6, 2023 at 9:53 AM Matt Mahoney <[email protected]> >> wrote: >> >> >> >> ... In the US, racial discrimination has been illegal since the 1960's >> and television has been portraying a colorblind world since the 1970's with >> no effect.... >> > This is the kind of thing that would be verified or debunked by Hume's >> Guillotine: >> > https://github.com/jabowery/HumesGuillotine >> >> 27,077,896 LaboratoryOfTheCountiesUncompressed.csv-8.paq8o >> ... >> 91,360,518 LaboratoryOfTheCountiesUncompressed.csv >> >> ... >> I am pretty sure that a program that found correlations in the data, >> such as between population, race, age, income, and crime, would >> achieve better compression. How would we use this information to set >> policy? >> > > Better compression requires not just correlation but causation, which is > the entire point of going beyond statistics/Shannon Information criteria to > dynamics/Algorithmic information criterion. > > Regardless of your values, if you can't converge on a global dynamical > model of causation you are merely tinkering with subsystems in an > incoherent fashion. You'll end up robbing Peter to pay Paul -- having > unintended consequences affecting your human ecologies -- etc. > > That's why engineers need scientists -- why OUGHT needs IS -- why SDT > needs AIT -- etc. > > The social sciences haven't yet come to terms with causality in a > *principled* manner. This is also at the root of AGI's troubles. Even > Turing Award winners specializing in AI causality, such as Judeo Pearl, > are confused about why Algorithmic Information is a superior model > selection criterion to progress toward discovering causal structures latent > in the data. > *Artificial General Intelligence List <https://agi.topicbox.com/latest>* > / AGI / see discussions <https://agi.topicbox.com/groups/agi> + > participants <https://agi.topicbox.com/groups/agi/members> + > delivery options <https://agi.topicbox.com/groups/agi/subscription> > Permalink > <https://agi.topicbox.com/groups/agi/T772759d8ceb4b92c-M73dfbfa0606d7918346871b2> > ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T772759d8ceb4b92c-M177d8045c70b9eb77542a522 Delivery options: https://agi.topicbox.com/groups/agi/subscription
