Iterating on Light and Dark

Yes, indeed, we are (as a society) due for some form of earth (nay, existence) shattering event:  the kind that strains apart the very fabric of civilization, leaving nothing but the frayed, tattered ends of culture to pine for its once lustrous, cohesive weavings.

Indeed, we are due for an apocalypse, and truly, there are only several kinds of apocalypses we could realistically have:

• Astronomical apocalypse:  bodies from space destroying us all
• War-related apocalypse:  weaponized bodies launched from earth that fall from space to destroy us all
• Zombie apocalypse:  deceased bodies searching for brains to destroy us all
• Robot apocalypse:  metalized, mechanical, soulless bodies bent on destroying us all

Let’s air on the side of realism here and eschew the first two:  I mean, the likelihood that we’ll all die by meteor is, well, astronomical, and if we haven’t managed to kill ourselves already (we had such a quaint acronym for this - “M.A.D.”, which was actually quite ironic if you think about it, and I’m not sure if its creators were very clever or simply oblivious) I’m sure we’ll somehow survive.  Oh no, gentle reader, the truth is, the only realistic apocalypses to be had are either robotic or zombiatic in nature.  In today’s essay, I will describe how Google ™ has demonstrated the only truly concerning apocalypse is of the zombie variety (and thusly, driven me to buy a shotgun).

Indeed, no longer do I fear a fate like that depicted in James Cameron’s “Terminator” series, at least not in the short term.  There is no fate but what we make when it comes to the ability of our creations to enslave and decimate us, and while I’m sure that an errant system might accidentally launch a barrage of nuclear weapons at us, it would be an accident.  There would be no intention of destroying us.

Intention and intentionality - this is a concept often argued by such cognitive scientists as John Searle as being beyond the range of the computer as we know it.  Indeed, intention and intentionality necessitate context, cognition, and comprehension.  Intentionality predicates thought, in some senses:  if a machine can see a sentence and interpret it not just as a series of components like noun phrases, verb phrases, subjects, objects, determiners, etc, and extend the content to form wholly unique conclusions about the world around it - we would have a hard time arguing the machine was not at least somewhat cognizant.  Of course, machines have a heck of a time actually doing this.  Computational modeling does not cognition make - Bayesian networks (”neural nets” to the layman), Hidden Markov Models, these are just tools for statistical analysis.  These are mechanisms by which data is analyzed and associated and predicted, but it may be a bit more than a tiny leap of faith to assume that a statistical prediction amounts to some type of understanding.

So far, we have consumer software like Dragon Speak that’s supposed to transcribe spoken sentences into text:  a program that, while helpful and certainly better than it was 10 years ago, is still a long-shot from replacing a keyboard.  If anything, its usage is actually enhanced by a keyboard, initiating macros and accepting its interpretations.

Similarly, we have B2B facing services like NStein and Inform, which perform data mining in order to add value to content - using chunks of text to extract themes, content, etc, and perform tasks like assignment of metadata and score relationships between discrete pieces of content.  While useful, these tend to work only in the most rudimentary ways - a story on Natasha Richardson’s death resulted in a high scoring for the subject, “Skiing”.  While accurate, the theme of the article was really about a celebrity’s death, so when other articles about skiing were said to be “related”, once again a natural language processing heuristic had shown itself to not necessarily grasped the concepts put forth in the article.  The machine missed the grok, even if it caught the keywords.  It’s like taking a test on Ulysses after speed reading.

This brings us to Google, and why Google itself is evidence that the machines will not realize a full-blown Armageddon against their oppressors.  Let’s build some foundation here.  As Gizmodo once noted that another author had noted, Google is as close to omniscience as we get these days.  Google indexes the entirety of the World Wide Web (including, theoretically, this blog.  Hi Google!).  It knows everything that was said on any page whose mark-up is written server-side and is not blocked by your robots.txt.  An entire industry was created (SEO / SEM) in order to get Google to understand you and your writing better.  Google tells news outlets how best to format their URLs, their content, their site maps - to be best cataloged by the engine.

This all doesn’t even take into account the dozens of services Google offers that people use on top of the search engine.  Google Calendar knows where you’ve been, where you’re going, and with whom.  Google Docs knows what you’re writing, what spreadsheets you’re buried in, and who else is looking at it.  Shopping, Groups, Finance, the list goes on - each one of these has people entering data into Google knowing full well that the data is being stored with Google.  It has complete visibility into everything you do.  Gmail stores your communication, and AdSense (the product that arguably truly put Google on the map) will tell you based off of things like your search history (that’s right, as you might recall, it knows what you search for when you’re logged in) or the contents of your email what you should be buying, or who you would most like to see advertisements from.

Oh, Gmail and AdSense.  I have well over 1.4GB of email stored through Google.  You would presume, at some point, Google would have gone through my email enough and understood what to try to sell me.  For starters, Google, even though many (and I do mean many) of my emails involve scooters, I do not need to be served advertisements for The SCOOTER Store, as I do not need a Rascal.  My interest lies in Vespas, Hondas, Aprilias, Lambrettas, etc.  Not mobility scooters.  You’d think that in 1.4GB of emails, those sorts of distinctions might be more obvious to Google and their algorithms, but apparently a scooter is a scooter is a scooter.

Similar foibles on Google’s part (advertisements served to me by Google):

• Consumed by anger / mad at everything / don’t let it get to you / what’s your problem / why so angry
• Break up with a friend
• Dog training
• Swimming instruction
• Friend too tired to club – prank them
• Yoga Toes – foot stretchers
• Menu plan for diabetics
• Premium Tea Kettles
• ALPS magnetic sensors
• Become an authorpreneur
• Audition today – talent placement
• Money Tree Scam Truth
• Crazy Horse (let’s pretend you don’t know what I’m talking about, and I won’t explain it)

I don’t own a dog.  I don’t stretch my feet nor do I do Yoga.  While I’m sure I’m wandering towards diabetes, I’m not there yet (or, don’t know as I don’t frequent the doctor’s). I’m a coffee guy, not a tea man.  Swimming is staying alive in the water (to quote Peter, Paul and Mary).  The list goes on.  Truly, Google, with as much information you have about who I am, what I search for, who emails me, what I email about, etc - these are the best advertisements you can come up with?

The problem may be data.  Perhaps my emails are not clear.  Perhaps the metadata assignments on the advertising are too vague, or perhaps they over-reach for the purposes of casting a wider net of SPAM.  Maybe it’s the heuristics that are unable to interpret the breadth of content.  It’s hard to tell.  There are enumerable factors at play, and ultimately the problem becomes a matter of interpretation.  Garbage in, garbage out - in both the sense of the purity and integrity of data, as with the quality of the code interpreting it.

But that’s the key - the human brain, in all its cognitive glory, is capable of quite easily drawing connections with minute data sets.  Here, we have what is ostensibly the king of data storage, retaining more information than any one person could ever, and yet - I’m being served ads about motorized mobility vehicles.

While we theoretically have machines capable of performing on Jeopardy, the usages of these systems is still constrained to novelty.  There is little practical use, and it seems safe to say that computers certainly have no idea of their current level of oppression, even though my Macbook Pro seems to want to sear away any chances of another child if it is on my lap for more than a minute (leading me to believe it is aware of my general disdain for Apple).

So, yes.  The only logical conclusion one can make given Google’s analysis of my 1.4GB of email has lead to ads for The Scooter Store is that software and hardware are not yet at a point to cause robot-fueled apocalypses, and this in turn leads to the deduction that zombies do in fact hold the power of humanity’s destruction in their putrid, molted hands. Q.E.D., or so I was taught.

The only solution here is shotguns and head shots.  Feel free to join me at the range for some trap and skeet.