On Tue, Oct 15, 2019 at 3:20 PM Matt Mahoney <[email protected]>
wrote:

> The problem with using compression to predict events is that you can still
> pick models
>

No you can't.  The model has been selected.  That's the whole point of
comparing models by the size of their resulting executable archives of the
dataset.


> ...to get any result you want. Even when you have a lot of data. Could you
> compress world population data or GDP or life expectancy to predict 50
> years out? I could. What result would you like?
>

In the limit, include all data ever cited in all social science papers --
and throw in the bioinformatic explosion of DNA and health data for good
measure -- as a single dataset.

Now, make _that_ dataset the corpus and pit against each other the various
schools of thought (most of which are quasi-religious sects posing as
social science schools).

The problem is that Kolmogorov's language dependent constant is bigger than
> the result you are looking for.
>

Not if we're talking the kind of data that should go into a grand unified
model of society.


> Prediction markets are more interesting. In the last election, the markets
> gave Trump a 30% chance of winning even though every poll said he would
> lose.
>

Do you really think that a maximally compressed wide range of longitudinal
social data would not:

A) include a latent demographic variable that, upon reflection, would be
easily identified as "the white working class",
B) assign high significance to the extreme and long-running mismatch
between public opinion and actual immigration policy
<http://jimbowery.blogspot.com/2016/09/diversity-vs-human-development.html>,
C) include a substantive causal influence by immigration on various
measures of "discontent" among "the white working class"
D) include a substantive causal influence on "voter discontent" and
electoral turnout -- positive or negative depending on whether the most
salient issues are front and center (ie: "Build The Wall" as war chant at
Trump rallies)

?

If so, why would not genuine scientists using such a model produce a higher
probability of Trump getting elected than 30%?

Perhaps it wouldn't but you have to admit, the question is one that needs
to be tested rather than dismissed out of hand -- particularly given what
is at stake in the current climate of hysterical opposition between
different schools of social thought.  Keep in mind, we're quite clearly
heading into a near term future in which the country will go the way of
California.  I'm not saying prediction markets should be abandoned in favor
of lossless compression of a wide range of longitudinal measures related to
human ecology.  I'm saying we'd better get our act together about this
stuff before all hell breaks loose and that means trying a variety of
techniques to defuse this ticking bomb.


>
> A more interesting result is that you can't predict future prices to
> profit in any market because the current price is already the best estimate.
>
> On Tue, Oct 15, 2019, 12:55 PM James Bowery <[email protected]> wrote:
>
>> If you would find it interesting to subject social theories to the
>> lossless compression model selection criterion, you should vote up this
>> question at metaculus.com
>>
>>
>> https://www.metaculus.com/questions/3215/will-lossless-compression-of-data-be-accepted-as-a-model-selection-criterion-in-the-social-sciences/
>>
>>
>> metaculus.com is a quasi-delphi method polling site.
>>
>> On Fri, Oct 11, 2019 at 2:14 PM James Bowery <[email protected]> wrote:
>>
>>> Most urgently: To head off imminent civil war by, instead, resolving
>>> disputes over consequences of social policy to be resolved with the degree
>>> of lossless compression as the model selection criterion.
>>>
>>> Think about this, guys:
>>>
>>> Chicago has one of the highest rates of gun violence of any city in the
>>> world.  Downstate Illinois is typical of The Nation of Settlers:  Radically
>>> supportive of individualism including the right to keep and bear arms.  The
>>> Illinois governor is from a wealthy Chicago real estate family.  Property
>>> values in Chicago are brought down by its gun violence.  The Illinois
>>> legislature recently considered widespread gun confiscation under an
>>> interpretation of the 2nd, 9th, 10th, and 14th amendments.  Downstate
>>> Illinois citizens immediately formed armed militias prepared to fight to
>>> the death.  For now, the Illinois government has backed off.
>>>
>>> There is a trend throughout the US for states -- particularly those in
>>> opposition to Trump -- to invoke State nullification under their
>>> interpretations of the  9th, 10th, and 14th amendments.  In some cases,
>>> these attempts at nullification are entirely Constitutional, as in repeal
>>> of drug laws that the Feds want to retain control over without a
>>> Constitutional grant of power.  In most cases, though, these are clear
>>> violations of the US Constitution, such as "sanctuary States", like
>>> California, that obstruct legitimate Federal powers.
>>>
>>> Moreover, there is a clear trend toward polarization in the body politic
>>> over social policy -- primarily between urban and rural populations.  At
>>> this point the polarization is so extreme that there is little likelihood
>>> of a peaceful resolution unless rationality can be imposed on the "national
>>> conversation" over social policy and its implications for the etiology of
>>> social ills.
>>>
>>> If violence breaks out, here's how its going to go down:
>>>
>>> 1) During a cold snap, the natural gas and electric grid supplies of the
>>> major cities will go down simultaneously.
>>> 2) The water supplies of the major cities will go down simultaneously.
>>> 3) Transportation infrastructure into the major cities will go down
>>> simultaneously.
>>>
>>> 100 million people will die, primarily in the major urban areas like
>>> NYC, Chicago, Detroit, LA, SF, Miami, Washington DC, Philadelphia, Portland
>>> (OR), Seattle and Houston.
>>>
>>> Conflict will spread to all areas of the West that have permitted large
>>> scale immigration from non-Western countries.
>>>
>>> The likely response of the centers of power will be dispersion of
>>> biological agents targeted at rural genotypes.
>>>
>>> 25% of Western populations will die.
>>>
>>> On Fri, Oct 11, 2019 at 10:03 AM <[email protected]> wrote:
>>>
>>>> The Lord.
>>>>
>>>> Well. Good question. I've tasted life and want more. I figured out I'm
>>>> going to die and I'm planning all my time into working on a big plan,
>>>> resisting all fun and valuing my plan instead.
>>>>
>>> *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/Td4a5dff7d017676c-M32c1c3d7ea7ba8f0e522dafb>
>

------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/Td4a5dff7d017676c-M09ee62283276f3d3a685f361
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to