The Future of Work

KIMO is focused on enhancing human learning through artificial intelligence. In this blog post, we’ll explain why we believe this focus is crucial, and why finding better technology to drive human learning is a necessity to drive human development for our future society. We’ll first explore the best estimates of changes in the job market in the coming decades, look at where we stand today, and end with what we believe will be part of the solution.


Job Markets: Expectations for the Coming Decades

To start our story, we have to move to Oxford, where researchers Frey and Osborne published a paper around the Future of Employment in 2013. According to their estimates, 47% of the US population would be at risk for automation in the coming two decades. It goes without saying that the results caused an outcry and a demand for more research in these areas. It is interesting to reflect on when Frey and Osborne made their predictions: 2013 was three years before AlphaGo defeated Lee Sedol. The AlphaGo ‘defeat of humanity’ was a game-changer, after which in particularly China strengthened their public commitments to be no.1 in the world of AI and investments surged. So, things only speeded up after their prediction.

When in 2017 other Oxford researchers interviewed the leading AI experts on ‘when AI would outperform humans on specific tasks’ (note: this would make humans redundant in those jobs), they found this the picture below. Although the estimates for specific tasks differ, it is obvious that the expectation is that the next 20 years will bring a lot of change.

Since the early 2013 Oxford report, other investigations on the matter have appeared that echoed the same sentiment. PwC predicted that 38% of jobs will have a high risk of ‘automatability’ by the early 2030s; Bain predicted that by 2030 employers will need 20-25% fewer people; McKinsey pointed out that many jobs could be partially automated and experts like Kai Fu-Lee believe 40-50% of jobs could technically be automated in the coming 2 decades. It is worth noting that the possibility of automation for 40-50% of jobs doesn’t mean we will have 40-50% unemployment, as there are other factors at play (e.g. regulation, markets adjusting to gig economy models etc.) – but it is clear that the world of work, and the inherent meaning of it, will change significantly in the coming decades.

AI is also a tremendous opportunity, both for revenue generation (sales, marketing) as well as cost-cutting (automation, optimization). Recent PwC report estimated the potential GDP increase due to AI to be 15,7 trillion (note that the global GDP today is estimated to be ~84 trillion, so this is an 18.6% increase). A big part of this value will be generated by Deep Learning, a paradigm that mimics human brains in how it makes predictions. The ‘neural nets’ in Deep Learning specialize in pattern detection in data through parallel processing of information – quite similar to how our brains are believed to work. For years, it was hard to run such parallel computation models on our (sequential) silicon chips, but times are changing. For small AI companies like KIMO, GPUs can be ‘rented’ in the cloud from giants like Amazon for a small fee.


Automation Today

Examples of where AI is already used to successfully automate processes previously done by human beings can be found in how Alibaba and Amazon ‘man’ their warehouses. 70% of the work in those warehouses is automated. Look a bit beyond the warehouse into the delivery of goods, and you’ll find that drone deliveries are now being tested to replace human drivers. The movies below demonstrate what I mean.

In my lectures around AI, I see different reactions to the trend. Fear levels for this trend literally range from 1 to 10 in the same group, with people perceiving different risks and opportunities. On the whole, however, fear correlates with knowledge in the world of AI. What we’re seeing today is not just more automation of manual labor, it is instead the automation of cognitive labor. How complex the cognitive tasks solve by AI are is changing on an almost daily basis. Creating art, making music albums, automating stock portfolios, drafting novels, even generating jokes are within the realm of current algorithms.


A Solution: Use AI for Human Augmentation

There is a catch to all this. This is where AI research meets cognitive sciences. The progression in the world of AI doesn’t match our progression in the world of learning. How we learn hasn’t changed much in the last decades. That brings the problem that the process of learning – in particular for adults – is a deeply ineffective process for most of us. We know we should do it more often, but time is limited, deciding what to learn is hard, the overview of the content area you’re about to study is missing and staying engaged amongst other things to do in life is nearly impossible (only 5% of people finish an online course). Most organizations realize they face a problem, as very few (~1%) of their employees are actively engaged in the learning management systems they employ. Boards have mostly ignored this for now, and through marketing phrases spun up by L&D departments like ‘continuous learning’ they have found a way to still feel comfortable in the situation. This is illusory. The harsh reality is that most people in organizations today don’t spend enough time preparing for the future – whether this concerns artificial intelligence, fintech, blockchain or other digital economy skills. People are just doing their good old job, even when they know it will not exist 5 years from now. This is not the best way forward for our workforce today – especially not for the younger generations. As a beacon of hope, the reskilling problem is recognized by some policymakers in the EU. This has resulted in some (future) EU funds to drive technological solutions for workforce re-skilling, starting in 2020.

I hope this short blog post provided some new insights into the current status, risks, and potential of AI. Although predictions differ, what is certain is that the future of work will be quite different than what we see today. What is also certain, is that AI will become part of the solution.


‹  back to blog
  • Udemy
  • Coursera
  • Youtube
  • Udacity
  • Medium
  • Datacamp
  • edX
  • FutureLearn
  • Pluralsight
  • Massachusetts Institute of Technology
  • Berkeley University
  • Stanford University
  • University of Pennsylvania
  • Harvard University