Just in Time Learning to Combat Technological Unemployment

By Topher Cliver

Technological unemployment is happening, end of discussion. In fact it's been happening throughout our history. What is up for discussion is whether advances in artificial intelligence (AI), robotics, virtual reality (VR), drones and cloud computing, to name a few technologies,  are going to displace workers at a faster pace than what we have grown accustomed to or proven ourselves able to handle. On one side of the argument are people that say this time is different than past introductions of new technologies and that the job losses will come so fast and across such wide range of jobs that many people will be permanently unemployed. The other side says this is a luddite fallacy and that while there may be spikes in unemployment, new technologies always create many new jobs and an inability to predict what they may be so there is no cause for alarm.

The case for tech unemployment being different this time is strong. Historically, in many countries, as economies lost farm jobs the people came to the cities where they found manufacturing jobs. A modern poster child for this is China. As the manufacturing jobs went away in the United States, it’s economy transformed into a services economy. It's happening again as we see service jobs on the cusp of being replaced by AI driven applications from automated call centers to pill dispensing robots. The real first massive wave of job losses will occur in the transportation and restaurant sectors. What makes this time different in that this is not a change followed by a wave of job loss and eventually new jobs, it is wave after wave coming faster and faster until we hit a tipping point.

Unskilled workers could always count on being able to be employed in retail or at the very least fast food. Automation is already devastating retail and the automation of fast food is at the very beginning. You may have seen the robots that roll out drinks on a assembly line at drive through windows or ordered via an app or one of those touchscreen kiosks many chains have toyed around with. It's pretty easy to imagine this progressing pretty rapidly to a fully automated restaurant experience open 24 hours with no workers calling in sick and consistent quality food pouring out. 

The loss of manufacturing jobs hit the middle class hard as those were the kind of jobs that someone with only a high school diploma could do and earn a solid wage. Truck driving jobs are like that too and they are about to take a hit from self driving truck technology. It will likely start on the long stretches between cities but soon enough if will be end to end. That same tech will end up putting all sorts of drivers out of work from package delivery to pizza delivery. The remaining taxi drivers will lose out again but not to human uber drivers but automated uber fleets that will, ironically, spell the end of the gig jobs aforementioned. There are many millions of these folks and the vast majority of them will not be able to take an online class on programming and get a new job.

The other side of the argument says this is a version of the Luddite fallacy to believe that the latest technological innovations will be different from all those in the past and permanently change the employment landscape. Technology always eventually creates more jobs than it takes and the new jobs are better than the old jobs so it is a win for all working humanity. To fear this change because we can’t imagine what the new jobs look like is irrational. It's not a huge leap to imagine stagecoach drivers becoming truck drivers but how would you have envisioned the millions of farm workers who lost jobs and moved to the city and did things like become gas station attendants or assembly line workers before either existed?

So at one extreme there will be more jobs made than lost so the challenge is more about preparing people for their next career or at least their next job until the next wave. On the other end of the spectrum, artificial intelligence advanced to a tipping point where AI is developing new and improved AI. Because of the computational speed of AI it becomes a positive feedback loop where it advances so fast and by such leaps that it creates a hockey stick graph of intelligence and leaves mere humans on the cutting room floor of the movie of evolution. There would be no job that wouldn’t quickly be done more efficiently by machines at this point. Of course even that scenario may not be fatal for the human job as we know it if the AI progress so far past us we become as ants are to us. Human don’t serve ants or do their work for them. 

So this brings us to what we can do about it. Universal Basic Income has been proposed by many as a social safety net for those who lose their jobs to automation but I suspect living on the poverty line one step above living on the street will be little consolation to people who once worked and took care of themselves. That's assuming western countries were to find the political will and funding to pass such a measure with an amount even at the poverty line. Maybe it will happen slower than we think and there will be time for new jobs to emerge things will work themselves out. Maybe it won’t.

I never took a course, at any level of schooling, on how to use a spreadsheet. I never took a formal training course and although I bought a book offering “advanced excel” I have to admit I never read it cover to cover as I planned. I skimmed and it referenced it a few times but it mostly just sat on a shelf. Despite this lack of schooling, formal education or training I am a somewhat advanced user. How can this be? Enter Google. I self learned the basics of excel and then after that, if I needed to do anything I didn’t know how to do I simply googled it and found step by step instructions in writing, diagramed or in videos. Call it just in time learning.

I work with full stack software engineers and this is how they work every day. They have a problem they want to solve or need to do something a better way they simply Google it. This doesn’t just happen in the office space, have you ever looked up a video on Youtube to walk you through putting together a piece of furniture or toy? Ever watch a video on how to play a board game or cook a meal? We do this all the time today already. What if instead of us having to self serve the just in time learning, we had an AI constantly looking one, two, three or more steps ahead and queuing up how to do each of these things? The moment we get stuck, or better yet, the moment before, an instructional video or overlay could appear  on whatever augmented reality device comes out of the ashes of Google Glass™.

We should be incentivizing R&D into training utilizing the same cutting edge technologies that will put workers out of work but in a use case of just in time training. Something like Google Glass™ overlaying objects in the field of vision with pointers, diagrams, video clips etc combined with an IBM Watson™ like subject matter AI collecting and analysing relevant information and providing instructions on the next steps. You could imagine someone who has never used a tool being walked through changing out a water pump on a car engine without knowing the first thing about cars or what they just did. A human would be the hands of AI at that point but they would also be getting an education as they went. 

It’s easy to say the scenario above wouldn’t happen because a robot could just be the hands of AI and that would eliminate of the limitations and high costs of having a human in the mix. In the movies this is where some super nerd invents an android that not only captures all the movement nuances of humans but also has the chassis of a terminator and the battery life of the Energizer Bunny™. You never see the generations of small iterative steps that got them there. You don’t see the floor mopping bots or the car washing bots and the millions of other focused purpose functions that robots would be tasked with doing before we get to a single android that can do it all. It's a convenient plot device to move a storyline but we are probably a long way away from that that and while the robots will take many jobs, just in time training will open many more up. 

Combine the dexterity and versatility of a human with the knowledge and resources of an AI and various sensors and input/output devices and you have that aforementioned Android now, without the Terminator hyperalloy combat chassis. This simple twist could bridge a twenty year gap before we have the movie androids. That would open up all sorts of possibilities for human work as the limitations of education, experience and really even intelligence are removed from the equation. The question becomes, would a human tolerate being told what to do all day by a machine?

For the relationship between AI in our ear and humans as AI’s hands to work, psychologists and sociologists and other ologists will need to figure out how to facilitate a tenable relationship. A super micro managing AI might feel like it's constantly nagging us and at the other end, letting humans do things a stupid way will reek of inefficiency in a way that is the equivalent of painful to an AI. Maybe the AI would need to be our buddy and say things like “They want us to do take out the garbage. I know you’ve done this before so just let me know what I can do to help.” That could be followed up with visual queues overlaid in the glasses like a highlight of where the new bags are stored and what the PIN is to get back in the door.

Humans like to feel they’re autonomous and and allowed to interject our creativity in what we do. It's hard to think of taking out the trash as an act of creativity but our combination of choices is creative and this experimentation is how we find out what works best while helping stave off boredom. An AI will have to allow humans to be human, but not too human. Finding that balance will be the key to people not taking a hammer to their earpieces. Finding out how to incorporate AI to help us be better rather than boss us around with the goal of perfection. Giving us information when we need it, direction when we want it and subtle hints and queues the rest of the time will become an art that ironically will take a human to define, at least initially.

Imagine an AI for work in you ear and maybe only occasionally being useful for the weather or your calendar or what room the meeting you are headed to is in. Maybe then it learns to dial into a meeting for you and even plugs in the code then transcribes the meeting and sends notes out to all the participants. Maybe you tell it what you are trying to do and it plugs the correct formula into excel for you. Maybe you just highlight text from emails, websites etc and it organizes it all into a wiki-like structure organized by topic and retaining the source’s metadata. Maybe one day you can’t imagine work without it.

Many jobs will get AI help. If it is a voice in your ear then hopefully its in the form of a helpful assistant and not a micromanaging tattletale supervisor. Hopefully we get just in time learning on the job and not learning on our unpaid breaks so you can keep your job. There will need to be legislation and regulation to ensure the rights of humans don’t come under further attack in the name of the bottom line. If it's a sellers job market because of a bunch of jobs like truck driver lost to automation, and humans start to become commoditized because the expertise lies in the software and not the wetware, then we may create an environment dangerously rife with exploitation.

So is a robot coming for your job? Most certainly the answer is yes but the estimated timeline for that happening varies a lot. If you work in a factory inserting the same screw in the same part over and over, well it's amazing that you still have that job. If you drive for a living then it might be a good idea to look at your options. If your job has a lot of repetition of any kind, maybe you should be worried. If you deal with people or work in varied environments or the problems you solve are constantly different then you may want to prepare for a digital partner, a google home with IBM Watson’s domain knowledge whispering in your ear. Maybe you’ll be doing something completely different like Roomba® repair and just about the time you start understanding it you’ll be doing something completely different again. Maybe the answers will come just in time.

by Topher Cliver,

-Author of the Novel, Plato's Dream: Crisis of the Employment Singularity.  

No comments:

Post a Comment