Showing posts with label deep learning. Show all posts
Showing posts with label deep learning. Show all posts

Saturday, 17 January 2015

We're Not Ready For This: A.I.

I saw this the other day:

He goes over deep learning, self-directed computer intelligence for the first fifteen minutes or so and summarizes at about 17:00 minutes.  The social implications of deep machine learning are quite profound.

Here are some other artificial intelligence related media that you might want to peruse:

Ray Kurzweil's The Singularity is Near (a long and tedious mathematical read with some wonderful implications mixed in.)

Her, Spike Jonze' ode to A.I..

A lot of Hollywood A.I. talk falls short into HAL type horror, but this one doesn't, it goes all the way.  By the end you'll be questioning our short comings rather than fearing what a superior intelligence might do.  I wonder what Kurzweil thought of the A.I. in this film and what it ends up doing.


Better education doesn't help? Work is irrelevant? What do we do
in a world of human pets that serve no real function in terms of survival?
This could be an age of unprecedented creativity, or the beginning of the end.
The TED talk has an interesting moment in those final two minutes where Howard is talking about the social implications of an imminent (the next five years!) machine intelligence revolution.  He talks about computers taking over jobs that we consider to be human-only and doing them better than people ever could.  This isn't about coding a better piece of software, it's about computers coding themselves in a never ending cycle of improvement.  It's also about people no longer having to be responsible for their own survival decisions.


What happens to insurance companies when automotive accidents are a thing of the past?  Accidents don't happen when the A.I. managing it can not only control the car in question, but also move the entire traffic jam up ten feet to avoid accidents.  This is often misunderstood as people say that A.I. driven vehicles could have bad code that causes a massive pile up.  These aren't machines running code, these are machines that create code as they need it, kind of like people do, but much faster, and with absolute precision.  And however well they do it now, they'll do it better tomorrow.


What happens to human beings when they are no longer
responsible for their own survival?
The busy truck driver still needs to sleep, what replaces him won't.  It'll never drive tired or hungry or angry or distracted either.  It'll only ever use the least amount of gas to get where it's going.  One of the tricky things about trying to grasp human superior A.I. is in trying to envisage all the ways that it would be superior.  That superior A.I. would never stop improving, it would take over any concept of efficiency in business.

As Howard says, machines that are able to build machines in a continuously improving manner are going to make the social change caused by the industrial revolution look like a blip on the radar.

Perhaps the hardest implication of a machine intelligence revolution is the idea that your income is tied to your usefulness.  Our entire society is predicated on the idea that your income somehow reflects your usefulness.  If human usefulness is no longer tied to social status, what would society look like?

During the big market bailouts in 2008 someone online described business as the cockroaches that feed off the work of human society.  He suggested that you don't feed them steak, you just let them thrive on the waste.  The implication was that capitalism is a necessary evil that serves human beings, not the other way around as it's often stated (people are a necessary evil in capitalism).

The idea that people could be free to pursue their own excellence in the future without having to work for the cockroaches is quite thrilling, though it would require a huge jump in social maturity for human beings.  We'd have to begin identifying our own self worth through our own actions rather than our education and employment.  I suspect most people aren't close to that.  We'd also have to recognize that everyone has a unique and valuable place in society, which sounds like socialism!

Education is as guilty as any social construction in aiming children towards the idea of success being employability and income.  We stream students according to their intellectual capital and then tell them to work hard in order to achieve financial success in the future.  The very idea of effort is tied to financial success - something we'd have to change in a machine intelligent future.  Can humans value themselves and seek excellence without the yoke of survival hung around their necks?

Universal income is an idea being floated in Switzerland and elsewhere.  If the future is one where people are no longer integral to their own survival, we better find something other than a survival instinct to base our self value on, or we're going to quickly run out of reasons for being.

Wednesday, 21 December 2011

Photoboarding

Thanks to @fitfatman I now have a working term for an emergent student behavior:

PHOTOBOARDING: an emerging student response of taking pictures of in-class notes from overheads or the board, rather than writing notes.

In a senior academic English class we began a unit on short stories.  The instructions were complex and specific (across several pages) and involved creating a lesson on the short story of their choice, and then teaching the class that lesson (good prep for university seminar work).

Pretty much every student looked over the paper without reading it, made no effort to create a plan based on the criteria and then talked about their weekend with each other (while occasionally complaining about how much reading was involved with this unit).  It's a week before the holiday break, they weren't particularly motivated to be there.  Fortunately universities never set exams or anything important right before the holiday break.

One of the sharp ones came up and asked for clarification.  I spent 20 minutes sketching out a timeline/chart based on the criteria in the assignment with him on the board (in other words, I made notes).  By the end of our chat he had a clear idea of what to do on this specific assignment (I didn't suggest anything, I simply wrote down what he found in the handout).  He also had a useful means of organizing himself for future assignments.

When we were done half a dozen students came up and took photos of the board with their phone cameras.  There were maybe 50 words in the chart.  In talking to other teachers, this appears to be an emerging student habit, taking pictures of notes written in class.

There are a couple of difficulties with this.

It turns out that writing by hand creates all sorts of interesting neurological connections between the sense experience of hearing and seeing, and the development of memory structures around new ideas.  I'm guessing that the 'push a button' approach doesn't create the same linkages, and doesn't allow you to work through the material a piece at a time so much as simply grab it up all at once, and they aren't even the ones doing the grabbing.

The other difficulty lies in what this approach says about what happens in a classroom.  Students often come in after missing a class and ask what they missed.  They expect access to information, easily handed over, often online.  If information transmission is all that happens in the classroom, then you really don't need a teacher to do that.  If information transmission is the point of education, then we really don't need many people at all.

In the moment that I modeled, experimented and tried to demonstrate a self-sufficient way for a motivated student to get a handle on complex instructions, I wasn't merely transmitting information, I was trying to create a memorable moment using written and verbal cues that would give him the tools to deal with this situation in the future.  The notes were an expression of this, but the goal was a change in his behavior that allows him to be more self sufficient and effective in dealing with complex tasks.

Taking a picture of the notes, reducing that moment of teaching to a few pieces of information on the board, fails to recognize the importance of internalizing learning.  If we develop digital habits that limit our ability to effectively remember what happens, and in the process reduce the complex internalization of ideas by simplifying teaching and learning into information transmission, we're one of the main components in the creation of digital natives who wallow in the shallow end of learning.

Many teachers speak of their students' horrific memory.  Without the process of deep reading and writing to gradually introduce ideas into our minds, we become surface dwellers, never considering ideas in deep, contextual ways.  Our brains are able to consume great amounts of detail if the information is streamed in (reading and writing happen to do this wonderfully well); snapping a picture does not allow for that.

The mechanics of reading and writing aside, my real concern is in the externalization of ideas.  It is going to become increasingly difficult to teach (encourage growth in understanding and resultant behavior change) if the process of learning is simplified into data transference.  In courses of study that develop complex curriculum over long periods of time (ie: all of them), we are displacing complex neurological actions that develop deep, contextual understanding and provoke personal growth with the click of a button.

As long as technology is seen and sold as a means of simplification and way of reducing effort, we're doing our students a disservice by pushing it, and ultimately creating imbeciles.  Until we begin to advocate for technology that doesn't dumb us down, for technology that allows us to effectively complicate and empower our thought processes, we're part of a major societal problem.

In the meantime, students take photos of notes, replacing a cognitively engaging means of remembering and internalizing new ideas in a personal, contextual manner with the push of a button jpeg.  Most digitally interested teachers would call this efficiency, but we seem to be constantly confusing efficiency with gross simplification.

Why We Remember What We Write
http://www.lifehack.org/articles/productivity/writing-and-remembering-why-we-remember-what-we-write.html

Memory & Writing
http://www.neurology.org/content/46/5/1467

What is learning?
http://www.learningandteaching.info/learning/whatlearn.htm

Deep Reading & Memory: A professor's guide to integrating writing...