How Laura Poitras Helped Snowden Spill His Secrets - NYTimes.com

[Steve Smith]

Marcus

>> This sounds a lot like the problem of verifying computer-generated
>> proofs like the early example of the 4 color problem.  It might be
>> almost good enough to be able to verify each "step" of the proof and
>> the "logic" that it all hangs together with, even if no human can
>> claim to actually intuitively grasp the entireity of it?
>>
> Not a constructive proof, I'm claiming that from a bunch of wonky
> premises the `candidate for proposed trust mode 3' can iterate a
> argument forward in a useful, convincing, or subjectively interesting
> way that causes me to listen for more.

Good point... though I *do* think that law purports to do something
similar to a constructive proof?
> It might just be because they're a good at sales and that I'm a
> sucker.  The Kool-Aid is tasty, may I have another glass?!
I don't think you were born when that image was coined were you? Ok...
probably (1978) ...   In a strange twist of fate, it turns out that
"Kool-Aid" gets the questionable "credit" for Jonestown, but the actual
flavoring was "Flavor-Aide", a lesser known flavoring like "Kool-Aid".

- Steve

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com

[glen ropella]

Roger Critchlow wrote at 08/20/2013 04:51 PM:
> An obituary for Elmore Leonard noted that his characters, when stressed,
> would suddenly act.  But you never knew whether they were going to solve a
> crime or commit a crime or both at once until the act played out.

This adds a layer on top of TT3.  It strikes me that the good authors can make their characters 3D by adding "just enough" surprising behavior.  But trust is a relationship between 2 parties, which means each type has to be understood in terms of both its domain and co-domain.

TT1 (dist from Truth): many approximations to one Truth
TT2 (subject estimates the object): many subjective perspectives to many estimates of the object's attributes.
TT3 (journey not the destination): one subject's criteria for interestingness to many individuals who cover subsets of the criteria
TT4 (empathy): one subject to one object
TT5 (canalizing commonality): one predicate (e.g. alleged mobster) to one satisfying property (e.g. committing mobster crimes)

Leonard's reader-character relationships could be thought of as either TT[23].  But, I suspect TT3 is the better way to describe it.  That's the point of the emphasis when people say _narrative_.  TT[25] are often used as devices for sit-coms, where narrative is de-emphasized in favor of stereotypes.

--
⇒⇐ glen e. p. ropella
The first ones to sizzle on the judgement day
 

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com

[glen ropella]

Steve Smith wrote at 08/20/2013 07:18 PM:
> I depend on what you say about the way it works, but that doesn't stop me from seizing up when an opinion or decision is handed down that is "just plain wrong".  Lawyers, i suppose will try to trace back it's provenance to find where the flaw occurred.  Me, I just want to hold a mirror up to, or shine a light on, it's flaws.

I think this is the heart of the problems many people have with the speculation that computers (as we know them) are sufficient for generating consciousness.  I seem to remember Penrose making the argument that human mathematicians can "leap" to proofs (or methods of proof) that can't be found algorithmically (walked toward by purely mechanical means).  The same would be true of law as construction, rather than law as declaration.

At the end of the computation, if you can look at it and say "This is wrong", then you're effectively playing the domain expert in a face validation exercise.  It's tantamount to claiming "that machine can't or shouldn't produce that output".

I think the most interesting games are those immune to face validation, where either the declarations are impoverished compared to the constructions or the space circumscribed by the declarations seems much much larger than what the machine can construct.  If the experts can't tell whether or not the end state obtained through construction or (merely) chosen arbitrarily from the space of "legal" outcomes, then you have an interesting game.

--
⇒⇐ glen e. p. ropella
Roll up your expectations, and feed them into my sleep
 

============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com