Wednesday 20 April 2016

Deep Learning AI & the Future of Work

Most jobs have tedium as a pre-requisite.  No one does tedium
better than a machine, but we still demand that kind of work
for humans... to give them self worth?
This isn't the first time our compulsive urge to assign monetary value to survival has struck me as strange.  This time it was prompted by an article on deep learning AI and how machines are close to resolving many jobs that are currently reserved for human beings (so that they can feel relevant).  We like to think employment is what makes us worthwhile, but it really isn't, and hasn't been for a long time.

The graph on the right is from that article and it highlights how repetitive jobs are in recession as machines more effectively take over those roles.  As educators this leaves us in a tricky situation because we oversee an education system modelled on factory routines that is designed to fit students into repetitive labour (cognitive for the 'smart' office bound kids, manual for the other ones).



How can an education system modelled on Taylorist principles produce students able to succeed in the Twenty-First Century?  It can't, because it can't even imagine the world those students are going to live in.  There is a lot of push back in educational theory around the systemic nature of school administration, but I see little movement from management other than lip service.  Educational stakeholders from unions to ministries and even parents like our conservative education system just the way it is.
Between neuroscience and freeing ourselves of academic prejudices
(ie: creativity happens in art class), we could be amplifying what
human beings are best at instead of stifling it. (from Newsweek)
In the meantime, people who are taught to sit in rows, do what they're told and hit clearly defined goals are becoming increasingly irrelevant.  We have machines that do those very things better than any human can, and they'll only be doing more of it in the future.

Ironically, just at the time where human beings might have technically developed a way out of having to justify their survival all the time they are also crippling their ability to do what humans do best.  In recent years creativity,as critically assessed in children, is diminishing.  The one thing we are able to do better than machines is being systemically beaten out of us by outmoded education systems and  machines that cognitively infect us with their own shortcomings!

Machines offer us powerful tools for a wide variety of tasks.  I use digital technology to express my interest in the natural world, publish, and learn, but for the vast majority of people digital technology is an amplifier of bad habits and ignorance.  Many people use the personalization possible in digital technology to amplify their own prejudices, juice their brains like Pavlovian dogs in empty games, and all while living in a cocoon of smug self justification.

Just when we're able to leverage machines to free human beings from the tedium of working for a living, those same machines are shaping people to be as lazy, directionless and self assured as they wish.

In the meantime the education system keeps churning out widget people designed for a century ago and the digital attention economy turns their mental acuity into a commodity.

Rise of the machines indeed.





A nice bit of alternate future, but the description at the end is chilling - it's how I see most people using the internet: "At its best, edited for the savviest readers, EPIC is a summary of the world, deeper and broader and more nuanced than anything available before. But at its worst, and for too many, EPIC is merely a collection of trivia, much of it untrue, all of it narrow, shallow and sensational, but EPIC is what we wanted..."

Monday 4 April 2016

Dealing With The Impossible


Two decade old parts mean things don't fit together.
Making something work in this circumstance seldom
has anything to do with following directions
The other day I was trying to install carburetors on an old motorcycle (I was a millwright before I was an IT guy). I wasn't even sure if what I was doing was possible. I spent a couple of frustrating hours trying before I pulled it all apart and did it over a different way.

What I love about technology and engineering, especially when it involves free-form building rather than following directions, is that you have no idea if what you're doing is possible. This never happens in digital environments - they're all designed for you to eventually succeed. Kids think video game wins are wins, they're not, they're a conditioned response.

Any teacher who thinks free form building is just for fun is the kind of teacher who only wants students to perform conditioned response with a predetermined outcome (I'm guessing so they can control the situation). A lot of people (students and teachers alike) think that's learning. I think it's all about management and control, and it's one of the emptiest things we can do with students.

We shy away from stochastic processes in the classroom because we believe that failure is the inability to do something rather than an opportunity to better understand complex and open ended situations.

When trying to put together those carburetors I was unsure if the process I followed would lead to a successful outcome.  That uncertainty filled me with doubt and made me question what I was doing in a way that no lesson ever would.  We desperately hope for metacognition in student learning and then stifle it with overly restrictive learning goals.  No student ever starts a math problem, writes an essay or even plays a video game wondering if what they are doing is possible, yet most of the world, when it isn't a digital distraction or a lesson, works that way.  I suspect the cockiness I see in student attempts at engineering is grounded in the fact that most of their world (digital, educational, or worst of all: both!)  is a coddled exercise rather than a stringent test of reality.

In a classroom we like controlled circumstances with defined and plausible outcomes because they suit easy analysis of work completion, collection of assessment data and cement the teacher's place as the all knowing master of learning, but that limited circumstance doesn't offer much in the way of learning real world outcomes.

What would a learning environment look like if it wasn't modelled on data collection and teacher insecurities?