Where Mind Meets Message

It was early 2006. A cold chill still pinched my skin as I walked to Derby Hall on The Ohio State University campus.

In the third floor office of my friend, Ed Palazzolo, he unveiled his new tablet computer.

Dr. Palazzolo was, it turns out, ahead of his time. He exhibited it’s many features, and I was impressed. I even wanted one. But at the time, I could not come close to justifying the expense.

I thought of Ed often as Apple ushered in the era of the iPad. As I often do, I waited on the sidelines for early adopters to work out the kinks.

In Spring 2011, before the iPad2’s release, I was teaching Advertising Writing (ADV 3312) at Texas Tech. I needed a technical, high-involvement product for the students to write ads. They needed to learn about longer body copy.

I chose the Motorola Xoom(R), which for that day was superior to the first generation iPad. It no small bit of irony, it turns out that the iPad2 was released on the exact same day their ads were due.

Selling the Xoom was tough. It couldn’t be about emotion or image. Apple owns that. It had to be about facts. And ouch when Apple delivered on the dual-core processor and front-facing camera.

My interest was piqued.

True to his personality, my friend and doctoral student, Brandon Nutting, waited in line that first day to get that iPad2. Oh, sure, he has some “story” about how the timing was a “coincidence,” but I know better.

The turning point came one day talking with a student at the coffee shop in the student union. I was trying to explain something, and Brandon pulled out his now-dated iPad2 and loaded up a drawing program.

It was similar to one that Palazzolo demonstrated five years hence.

I started to see the possibilities. And although I adore my 17″ MacBook Pro, I have tired of lugging it and the power cord home every day when the most complicated thing that I typically do is edit a Word document or presentation slides.

And — this is the great excuse to trump all excuses — I teach mass communication, so I need to stay current.

We’re fewer than 24 hours into the experiment. This sucker was difficult to come by in Lubbock, as I had no desire for the 3G capable version. I ended up buying more memory than I wanted and had to drive all over Lubbock to find a black iPad.

Funny, I offered to buy my wife an iPad. She had no desire. Last night I was playing Words With Friends and she saw an ad at the top for the iPad Kindle app.

“It has a free Kindle app?” she says. “Take this iPhone back. I want an iPad.”

See? It’s paying off already. I teach advertising, and I got to watch the power of mobile advertising first-hand. Now where’s that tax deduction form?

{ 3 comments }

With all respect due Mr. Robert Fulghum, I think there’s a pretty important life lesson not learned in kindergarten.

Instead, I should have learned the most important thing I needed to know in the first week of freshman biology.

The concept is entropy, which Merriam-Webster defines as, “the degradation of the matter and energy in the universe to an ultimate state of inert uniformity.”

I’ve come to believe that this should be the guiding principle in life.

Whatever you build, no matter how beautiful or how tenderly crafted, will disintegrate into nothing.

This is the beauty of biological systems. Sure, we all die. That’s a bummer. But along the way, our cells perform a remarkable process of self-preservation along the way.

This last thought came to me this morning watching an episode of How It’s Made. Although I forget what was being made, I was spellbound by the speed and precision of a robotic arm.

Then I noticed the grease built up on what one could call its “elbow.”

And this precision machine in helpless to this buildup, and it will eventually succumb to it if some dutiful human doesn’t come along and clean it off.

Humanity has built some of the most elaborate social systems possible. They all fall. Romans. Gone. The Thousand Year Reich. Gone. (Thankfully).

The sandcastle you build on the beach will soon be destroyed. Soil erodes. The Parthenon crumbles.

Entropy. Destruction. Chaos.

It’s not enough to build something. You must nurture it, lest it, too, perishes.

This is true of the now rotten clothing of ancient people, the absence of which frustrates modern archeologists. It’s true of your automobile’s engine. It’s true of the societies we build, and it’s true of the relationships we build.

No matter how strong or how well built, they all just fall apart. Into nothing. Uniform distribution.

So take care of the machines in your life. But take better care of the people. Resist the altogether human tendency to take relationships for granted once forged.

For entropy is no kindergarten platitude. It is the law of the universe. And should you ignore it, one thing is certain: the chaos it brings will impose its will upon you.

Sandcastle photo courtesy of vpickering on Flickr

Parthenon photo courtesy of jscatty on Flickr

{ 2 comments }

If a college student asks me, “When is it acceptable to use a smartphone during a business meeting?” then I must confess that my only answer is “I have no idea … at all.”

Yesterday I offended a colleague by using my smartphone during a meeting, and it’s not the first time that this has happened. I respect and quite like this particular colleague, so I feel quite badly about the faux pas. However, the incident crystallized this question that has been floating in the back of my mind for some time now.

Before you merely label me a heathen with no sense of business etiquette, consider the facts of the case and a recent conversation with a highly successful alumnus of our program.

Students were presenting a strategic communication campaign, and their research raised a great question about the brand. The question. They cut to the heart of the matter. And sitting in the back of the room against the wall, I realized that their question was so good and so fundamental that I didn’t know the answer.

But I did know that the answer was merely a click away on my iPhone. So I set the phone on the notebook on which I was taking notes, Googled the question, and quickly learned the answer.

The volume of the phone was, obviously, off, and sitting on the notes on my lap, the physical movement of my Google search could not have been more animated than taking notes. And most people in the room were taking notes. So it could not have been a visual distraction.

Yet being on the phone was offensive in and of itself. As a behavioral scientist, I find this fascinating.

To me, information always is preferable to no information. And my data searching was subtle, especially considering that I was against the back wall behind the tripod videotaping the presentation. All eyes were, rightly, on the students.

Thus the faux pas was philosophical in nature.

I wish that I could claim complete innocence and report that as soon as I learned the answer, I pocketed the iPhone. I did not. I sent a couple of tweets about the presentation. Also discretely but less innocent, philosophically.

Industry perspective

My smartphone usage ruffles feathers more often than I would like, and because of this, the people who chafe at my usage likely would be shocked by how often I refrain from relevant searches due to feelings of etiquette.

All of this likely would merely heap shame upon me if it were not for a recent conference call.

As part of the college’s outreach, several members of our National Professional Advisory Board volunteered their time to talk with a group of advertising students about careers, success, and the industry.

When answering a student question about how to add value to an employer, one successful business-owning alumnus said that he especially values energized young employees who can put information at his fingertips when he needs it.

“When we’re in a meeting, and a piece of information will help inform the discussion, I need a person who can get that information. That person is invaluable to me, and I will always keep them near.”

OK, that is a paraphrase due to imperfect memory, but the sentiment is accurate, and the word “invaluable” is a direct quotation. And he elaborated that he was specifically referring to the employee who pulled out a smartphone and searched the information in the meeting, adding that he didn’t have time to do it because he was leading the meeting.

So out-of-the-blue, one of our most successful graduates tells a group of students that such information-gathering is invaluable, and no other member of the board on the conference call contradicted him.

Your guess as good as mine

So what to advise young people?

If this question actually were posed to me, I would hedge and say something like, “know your environment.”

I assumed that I knew what my younger-than-me colleague would think about such smartphone usage, and you know what they say about “assume.”

So I made myself look like a jerk, and I irritated someone whose opinion is valuable to me. That’s on me for not knowing the environment.

But that advice still sounds hollow to me. When I sit in Faculty Senate meetings and see our at-least-a-generation-older-than-me provost on his phone, I get it. Important stuff happens during a 90 minute meeting. He needs to attend to it.

However, it’s important to know that not only is my opinion not universal, it may be a definite minority.

I apologized sincerely, and I regret offending the colleague. But in the heart of the information age with the Millennial generation fully coming into the workforce, this issue must be addressed.

In the end, I forgot the first rule of strategic communication: know thy audience.

Photo published under a Creative Commons license from liewcf on Flickr.

Tagged as: Advertising, college students, technology

{ 1 comment }

Photo shown under a Creative Commons license from MethoxyRoxy.

Everywhere I look, it seems as if I find the work of cognitive scientist Jeff Elman.

Not his work, actually, but those things that his work explains.

I awoke too early this morning to rouse the kids and get them ready for school. In those semi-conscious moments before waking them, I scrolled through the new items on Facebook.

There I saw this status update by an old college friend, “Patrick wants to know why I can remember lyrics from a song I knew 20 years ago but can’t for the life of me remember what I just walked into the kitchen to get?”

Even in my dawn-hating haze, I thought, “Elman net.”

It’s a common thought, too. So many things make sense when viewed through Elman’s neural network structure.

I try not to reify the Elman net. I try to be a dispassionate scientist and not force the data into a hypothesis.

But his network is that darned elegant.

Parallel processing

This is not a short story, you see. It features many players, decades of research, and intricate concepts.

But it’s an amazing story.

The human brain represents the most complicated processing device in the known universe.

By any account, the human brain holds about 100 billion neurons.

Yet to focus on neurons is to miss the point.

Magnificent though they are, neurons are merely simple binary devices. They are either on or off. An action potential either cascades down their length (nodes of Ranvier, anyone>), or it doesn’t.

The salient fact is that a neuron holds no information. Neurons merely represent electro-chemical instantiations of the 1 and 0 binary code of the computer on which I type.

Billions and billions

Instead, everything you know rests in the connections among neurons, called synapses. And these are quite plentiful. Most estimate that an average adult human has more than 100 trillion synapses.

With this number, any one remains unimportant. But great insight came when we discovered that information is stored in the patterns of connectivity.

The brain stores information by altering the connections among neurons, somewhat like traffic patterns in metropolitan areas are better understood by intersections than isolated stretches of highway.

Slow moving vehicle

Although neurons number in the billions and synapses in the trillions, remarkable limitations exist on these information-processing devices: they’re slow.

Just ask an electrician to design an electrical system submerged in water, and you’ll understand the challenge of wiring a brain bathed in cerebrospinal fluid. Conditions are, shall we say, suboptimal.

In part as a solution to this problem, evolution engineered the chemical transfer of signal of action potentials across a synapse via neurotransmitter.

Problem solved, but electricity moves near the speed of light. Action potentials lag behind, and the chemical transfer of neurotransmitters is far, far slower.

Without a clever solution, we’d be remarkably slow, never realizing a tree is falling until our brain was too flat to process anything.

The solution is a massively parallel system. A “thought,” as it were, does not travel down a single neuron. Instead patterns of signals cascade across networks.

Think of it this way: if you want to get 100 marbles somewhere in a hurry, it makes no sense to push them one at a time through a narrow pipeline. Instead, push 100 marbles at once through 100 narrow pipelines. They arrive together, far more quickly. Such is the difference between serial (one at a time) and parallel processing.

Building models

Once we understood this, cognitive scientists began to model this process, perhaps most famously by David Rummelhart and James McClelland in their groundbreaking two-volume Parallel Distributed Processing (1987).

Without going into detail, these neural network models can solve amazingly complex problems. But in their original versions, they have no concept of time. And time matters. A lot!

People began to try to build time into these models in a meaningful way, including the ironically named Michael Jordan (not that one).

You see, simple neural networks can learn a lot, and they can learn connections between pieces of information that most other types of learning simply cannot learn.

But it appears that they can never (to date) learn everyday things, such as grammar. Cognitive science and linguists share a tight bond, and if neural networks were to explain human cognition, then they had to make a reasonable prediction about completing the sentence, “John walked through the ___________.”

You want to say, “door,” most likely, and you would have been surprised if I had said “octopus,” and you likely would have been even more surprised if I had said “walked,” for the first is highly unlikely, and the second is not grammatical.

But you should have been far less surprised if I had said, “red,” as it is common for modifiers to precede nouns in English.

“Door” was most likely. “Archway” would have been OK. “Red” also unexpected unless you knew the door about which I spoke. But “octopus” and “walked” are out.

How do you model this?

Finding structure in time

Enter Jeffery Elman and his brilliant 1990 article, Finding Structure in Time from the journal Cognitive Science.

Without delving deeply into the architecture of neural networks, Elman made a very elegant change to the simple feed-forward neural network.

You see, the latter take some representation of the worlds (not unlike how your retinas turn light into neural signals) and process them into some kind of “output.”

So, for example, imagine seeing the letter “A” and knowing it is an “A.” In this case, light has been turned into an output of letter recognition.

Simple neural networks excel at this task, even learning to recognize the letter “A” through all the various different fonts, many of which are not A-like at all.

But what comes next? What is the most probable letter to come after an A? Or more importantly, if you’re reading along a string of type such as this, how do you know when an “A” signals thew end of a word? In English, words don’t typically end in “A.” But how do you actually know this?

Elman had a solution. You see, between the “input” and “output,” a simple neural network must have an internal representation. That is, it must be allowed to transform the “input” into some other, hidden, form in order to best produce the output.

So, to solve any meaningful problem, neural nets have hidden layers. There, math drives a seemingly magical process of turning the world into some other unique code needed to “understand” it.

Elman’s solution was elegant. Give the neural network access to its internal representation from the immediately preceding timestep.

This is like a short-term memory, but it is exceedingly short. One moment in time. So, for instance, if the network were processing this very sentence, it would have access to the “S” in “So” when it was processing the “o.”

But not quite. It would have access to its own internal representation of the “S” in “So” when processing the “o.”

One solitary moment in time.

But the use of these internal representations has dramatic results. Consider the processing of “The quick red fox jumps over the lazy brown dog.”

When you start, you have nothing. Your mind is proverbial blank.

Now: T Before: (nada)
Now: h Before: (T)
Now:e Before: h plus my internal representation of T
Now:(space) Before: e plus my internal representation of h (which included my internal representation of T)

So what happens? You build a short-term memory on the fly. It never actually exists more than a moment in time, but the information makes it stretch back much further.

Back to the Cabaret

Although it is hard to do it justice here, these networks are so incredibly powerful. They explain so much.

Because at each moment in time, a network that has “learned” about the world, expects certain things to come next, and it “knows” when the next thing is out-of-place.

For instance, many words in English end in “-ng.” Walking, running, typing. We have no problem pronouncing it.

But have you ever watched a native speaker of English try to pronounce Vietnamese names, such as “Nguyen”?

It’s funny really. Their faces usually contort trying to figure out how to begin.

You see, in English, words never begin with “Ng.” Ever. So the Elman net inside your head doesn’t know how to begin without the rest of the word to get it started.

It’s the same sound, largely. You say it dozens if not hundreds of times a day. But time matters. And “-ng” comes later rather than sooner.

Patrick’s question at last

In many ways, music is special. But it’s highly ordered. That Scorpions tune that Patrick hears in his head has played in his ear hundreds of times.

And every time, there was a very rigid structure to it. It precedes in the same order, and with small exceptions for live versions, etc., it’s always the same. Singing one line cues the next. One note cues the next.

It’s a well-rehearsed pattern, and Elman nets are exceptionally well versed at this, pardon the pun.

So the song is well stored, and any cue to it is highly predictive of what comes next. Once you get it started, it runs itself in memory.

Conversely, however, I am willing to bet that Patrick would flounder and stumble trying to reproduce the words to that song backward.

What about the kitchen? Well, Patrick has likely walked into the kitchen hundreds of times, too. Usually along a similar route.

This, too, gets stored as a memory in time. I’m guessing he never abruptly turns and smacks into the wall when alcohol is not involved.

It’s an automatic process because it is so well learned. But unlike the song, something different comes at the end of the loop.

He does something different almost every time he walks into the kitchen. So once you get there, the loop ends, and it has no idea what comes next.

En route, he engages in a well rehearsed automatic process. It’s so easy that it frees his mind to think of other things. The mind wanders to his plans for the day, and the immediate goal slips out of mind because it is not needed for the automatic process of walking. When he get’s there, short-term memory has been cleared, and the long-term loop is trying to guess the correct next step from the thousands of past results, none of which is accurate.

Think of it this way: you’re pretty good at predicting that next word in the Scorpions tune, but you’re pretty lousy at predicting the next song on the radio.

Tagged as: brain, cognitive science, music

{ 7 comments }