?

Log in

No account? Create an account
The search for search answers (2 of 2) - Baxil [bakh-HEEL'], n. My Sites [Tomorrowlands] [The TTU Wiki] [Photos]
View My LJ [By Tag]


November 29th, 2006
04:36 pm
[User Picture]

[Link]

Previous Entry Share Next Entry
The search for search answers (2 of 2)
To: "Fang, Claw N" <redwolves@therimail.com>
From: "Technical Support - FeralNet" <cbf@feralnet.net>
Date: Thu, 26 Oct 2006, 21:54:02 GMT -0800
Subject: FWD: Re: That search we were talking about


And here it is.

-- Forwarded Message -------------------------------------
To: "CB Fox (Feralnet)" <cbf@feralnet.net>
From: "George Sinclair" <geosincronous@dwimail.com>
Date: Thu, 26 Oct 2006, 18:16:34 GMT -0800
Subject: Re: That search we were talking about


> > A mystery indeed! Do you mind if I do a
> > little digging and blog about your friend's
> > search?
>
> Should be alright :) But just in case, maybe you
> should take his name out of it entirely, and any
> context for the search? I know he wants his
> privacy.

Fair enough. Here's what I posted.

Subject: A dwivination Discovery
Tags: dwivination, tinfoil hat, current events

Regular readers of our dev-team blog already know that one of our favorite games here at bwim is trying to puzzle out "dwivinations," those search results that are so counterintuitive or bizarre that even magic seems insufficient to explain them.

Another stellar example comes to us today by way of a friend of mine who just got hired to dwim-dev. (Congratulations, and see you in the office!) He related to me a search that someone shared with him regarding the New Atlantis Project's recent run-ins with the U.S. government. Take a look at totp://dwim.mag/search?type=saved&user=geosincronous&id=f55HX2aOTp95dB and tell me that didn't send your jaw to the floor.

For those of you following along without magitech or who are too lazy to click on the link, that would be a dwiv for "Why isn't the government magically searching for the Discovery?" (the Discovery being the NAP's flagship on their island-raising mission; great reading on them in the current Vanity Parade.) The top link goes to www.annapolis-usna.edu/journal/archive/2006-09-27/06jackson.html (subscription wall; BugMeNot). In a nutshell, a September news brief about a history professor being injured in a natural gas leak.

As others have written here, us non-digital beings can easily create ways to link two seemingly unrelated events -- albeit implausible and/or uncomfortable ways. We just don't expect our computers to have that same power. And they don't; search results follow a strictly logical and verifiable process. While DWIM's searches are based on proprietary magical technology, that magic serves exactly two well-defined purposes: Organizing the end user's thoughts, to distill vague desire for knowledge into a coherent question with a coherent answer; and comparing that question against a database of human-generated knowledge.

It can't know anything we don't already know; it can't play the what-if games that lead us into conspiracy theories. It can only search for and connect public facts. The chain that DWIM follows unrolls from question to partial answer to refined question to better answer, etc. It can't go anywhere the facts don't lead.

So how did it make the seemingly conspiratorial link between the continuing inaction of the world's most powerful government and a single barely notable pipe failure? As usual, the results of dev-team research are below the fold, to give readers a chance to scour the Web themselves before seeing the "official" conclusions.

<blog_fold>

Here we go, then. The first step to puzzling out virtually every dwivination is to click the "Related/Relevant" link (the small plus-sign icon) next to the top result.

"Related/Relevant" is the closest thing that DWIM has to a "Where'd this come from?" button. Due to the formidable difficulties of digitizing concepts, the magic-driven DWIM engine can't store the state information for the chains of facts it lays down between question and answer, but it does store what it can. When one of its chains of sub-searches terminates at a result with maximal local relevance density (a "node"), and when that node is marked as having been used to advance the main search chain, it stores information about that node. R/R scours saved nodes and ranks them based on relevance to the final result, which is a decent approximation of its "thought process" for the last several steps. Beyond that, there's too much chaos in the system to reconstruct anything.

In today's case, though -- as with the most fascinating dwivs -- there's nothing that leaps out at us. A diary post from the daughter of one of the other accident victims; a government document-dump from the scandal over Nigerian yellowcake uranium that generated so much attention in the run-up to the Iraq war; minutes from a Washington, D.C. veterans' group meeting. Huh? None of the top three R/Rs seem to have anything to do with each other, and only one is obviously connected to the original result.

Of course, that's on the surface. DWIM doesn't care about how obvious a fact is, only that it can find it. Closer readings of all of the page links, and the original article, begin to show glimmers of light.

A careful perusal of the veterans' group minutes shows a note the secretary inserted near the bottom: "Next wk Sun sted Sat, Navy using room (meet re Atlantis)." An interesting footnote, probably passed on as gossip that seemed innocent at the time. They do use a meeting room at the Douglas MacArthur Memorial Building, after all, and it's not like the government is in the habit of working weekends.

The yellowcake document doesn't have anything quite so relevant -- until we skim the diary entry and realize some names are familiar. Terry Rhodes, the father of the writer, and fellow accident victim Frederick Manchiewzski (mentioned by first name in the diary entry, but fully ID'ed elsewhere in her archives), are both among the group of CIA surveillance mages whose cover the Bush administration "burned" in the yellowcake scandal. The R/R document is the one that identifies them by name as agents working for the government in international magical surveillance.

News articles further down the R/R list confirm that they were reassigned after that incident -- but apparently remained in the magical surveillance arena. One unusually candid quote from an anonymous CIA official sums it up: "Now that their identities are public, they're safer with us than out on their own. Besides, their skills are too valuable to abandon."

And then there's the diary itself. The writer says she learned about the incident involving her father from a friend who was walking by the building at the time: "The window of the meeting room thumped like a bass drum, bulletproof glass turning into a white spiderweb of cracks. Half a block of pedestrians glanced up, then continued about their business or stopped to gawk. Three minutes later, the ambulances showed up, and she called me on her cell phone."

What's most interesting about this is a subtle detail from the original news story: A Department of Defense spokesman said the gas leak that caused this was from "an underground pipe." Yet pedestrians "glanced up" at a window facing the street?

When DWIM runs across an inconsistency like this, it's unwise to make a spot decision as to what the truth is. The dev team did the best thing we could think of: The search program seeks independent verification of the fact in question. If a preponderance of search evidence, or one of the few sources we label authoritative, backs up one side of the story, it goes with that. In some cases where the sources still disagree, it can weigh the motivations of both sides -- for instance, the word of an independent witness is judged more reliable than that of a participant, who has more self-interest in their story. Failing that, it treats both assertions as false.

In this particular case, verification was available. In fact, some verification comes from one of the sources we've already cited: the veteran's group meets in the building's second floor. (And their subsequent meetings have been inexplicably moved to a nearby Congressional office.) Building plans, near the bottom of the R/R list, show that what gas does come into the building doesn't even get within 50 feet of the meeting room!

Suddenly, this "gas leak" looks rather different. The source of the explosion is now an open question. Our professor -- who by the original news story's own admission was "chair of the school's [short-lived] Department of Magical Studies" -- and two known spy mages were among the attendees of a meeting a source identified as being about the New Atlantis Project. (The government says it was a meeting on Naval Academy outreach, but its credibility has already been ranked down by the verified lies noted above.)

One of my next questions, and apparently DWIM's as well, was what did cause the explosion. Digging a little further down the R/R list, a conspiracy Web site posted a copy of the police report filed after the blast. The cops only got a few details before the Department of Defense took over the investigation, but there's much to think about in their reporting that several of the building's thaumometers registered spikes at the time of the explosion, and others had been disabled entirely.

Now, there are some things that we still don't know. There's no evidence that the meeting was specifically a spy session. (Personally, I'd bet it was, but DWIM doesn't gamble.) There's also no evidence that the New Atlantis Project was involved. But then, there doesn't need to be. The picture developing is already pretty clear: A government mage meeting on the NAP was explosively disrupted by a magical surge.

Granted, the facts supporting this interpretation aren't overwhelming. Some of the links are circumstantial or tenuously sourced. But DWIM's job is to return the best answer possible; in the absence of well-documented governmental decisions, and in the absence of official statements on the subject (the Bush administration has had plenty to say on the NAP, but the subject of magical surveillance of the Discovery isn't exactly drawing big press conferences), it has to find the link with the best available evidence. In a void of conjecture and propaganda, this incident is both relevant and grounded.

Readers, feel free to flesh out the research or post questions below. As always, we welcome your dwivination submissions at dwim_dev@dwimail.com.

- geo

Current Location: ~calorg
Current Mood: accomplishedwriteiful
Current Music: "Secrets," Mike Oldfield
Tags: , ,

(2 comments | Leave a comment)

Comments
 
[User Picture]
From:dreamflow
Date:November 30th, 2006 12:52 am (UTC)
(Link)
-- bwim Reader Comment: The_Other_Dennis

Geo,

Fascinating analysis as always. This is why I love bwim. :) I'm still trying to find a good dwivination of my own to contribute. It's such a delicate balance of esoteric subject and available documentation.

But in the meantime, you missed a significant point. You wrote "There's also no evidence that the New Atlantis Project was involved." You linked to an article yourself that shows otherwise!

Yanking an excerpt from Vanity Parade's feature on the Discovery:
"Does that job include fending off magical surveillance?" I ask. "I can't imagine the Navy's only trying to track you visually."

"Oh, it's not just visual," he replies, ticking off a list on his claws. "Sonar, radar, picking up our radio broadcasts. A sub followed us for a while based on engine noise ..." Ben trails off. "Magical monitoring. I, er, I probably shouldn't talk about that. Let's just say they haven't tried it lately."

Looks like somebody has something to hide!

Posted by The_Other_Dennis, Oct 26 2006, 8:42 PM
[User Picture]
From:baxil
Date:November 30th, 2006 12:55 am (UTC)
(Link)
-- bwim Reader Comment: geosincronous

I actually did notice that, and thought about mentioning it in the article, but it doesn't show up on the R/R list and therefore wasn't a factor in result selection. Here's why. "I probably shouldn't talk about that" doesn't imply guilt, it implies knowledge. You and I can draw conclusions about whether that knowledge was firsthand or not, but the DWIM engine's job is not to speculate.

DWIM almost certainly ran across that article during its search, but the fact that a Discovery crew member appears to know something about the government's hesitation was insufficient in itself to advance the search chain.

Posted by geosincronous, Oct 26 2006, 8:58 PM
Tomorrowlands Powered by LiveJournal.com