The short answer to the above question is ‘lots of them’. But precisely which ones – and, more importantly, the extent to which that might impact on your prospects as you enter the workforce – isn’t nearly as straightforward.
For one thing, the idea of robots (or, less exotically, machines) performing parts of our traditional jobs isn’t new at all. They’ve been doing it since the early 1990s, in fact, when we first developed technologies that led us – and our computerized helpers – into the controversial modern field of machine learning.
Back then, these smart machines were mainly deployed to cover very basic, mundane, repetitive tasks that were generally seen as an irritating time drain in various common careers. By teaching machines to ‘read’ handwritten addresses, for example, we could put them to work sorting mail for delivery at the post office, or churning out initial risk assessment scores based on tick-box applications to insurance firms.
Rise of the machines
At the time, handing off these sorts of tedious admin duties to robotic assistants was widely seen as a great boon for employees, freeing them up to focus on aspects of the job that required more critical thinking or creative problem-solving approaches. The worry a lot of people have today, though, is about how successfully we’ve continued to develop these artificial intellects over the past couple of decades.
Today, we’ve got robots and algorithms that can use much more complex versions of the same basic ‘learning’ process – in essence, developing the ability to make judgment calls, by mapping new information on to vast databases of previous examples – to perform far more exacting tasks. Smart machines can already use this type of artificial intelligence (AI) to diagnose a wide range of diseases, play the stock market, and even grade student papers at least as reliably as (and often outperforming) their human counterparts in medical, financial and educational roles.
As humans, the long-term outlook for our ability to rival automated algorithms on these sorts of playing fields is, frankly, not great.
In fact, during a fascinating a TED talk in February last year, Anthony Goldbloom – the CEO and co-founder of data science and machine learning platform Kaggle – declared that humans have literally “no chance” of even competing with today’s machines in the successful completion of repetitive, high-volume tasks. And, as the technology quickly becomes more affordable, this will inevitably lead to significant (and often painful) upheaval across a great many job markets.
Data is the difference
Where machine learning can’t match human performance, though, is when it comes to dealing with what Goldbloom refers to as “novel situations”. Ask a robot or computer to get from A to B with a good enough road map – which is effectively what huge datasets provide – and they’ll beat us to the finish line every time. But, take away that road map, and there’s no telling where they’ll end up.
In other words, even the smartest machines aren’t currently very good at thinking on their feet, or coming up with creative solutions to problems they’ve never tackled before. Crucially, they’re almost completely unable to employ processes of deductive reasoning, meaning they can’t import solutions from one set of experiences and use them to solve problems in an entirely unrelated field.
By way of example, Goldbloom recalled the invention of the microwave oven, which came about after a WWII engineer (Percy Spencer) noticed that the radar unit he was working on had melted a chocolate bar in his pocket. That, according to Kaggle’s CEO, is precisely the sort of intuition that smart machines aren’t even remotely capable of at present; moreover, we’re not even close to understanding how we might develop any sort of AI that can.
While Spencer’s accidental ‘eureka moment’ might seem an extreme example, it’s actually based on a kind of creative deductive reasoning process that – on a smaller scale – we as humans instinctively perform many times every day. In terms of our long-term employment prospects, that’s huge for us, because plenty of roles call for a lot more of that type of behavior than you probably realize:
- Directly customer- or client-facing jobs. This is arguably the most obvious area where we’ve still got smart machines licked for a while yet. Any job that requires regular interaction with other humans – be they customers, clients, or even working closely alongside colleagues – are entirely dependent on a complex code of social interactivity that requires emotional intelligence, cultural sensitivity and a huge range of reactive flexibility to carry out successfully. Algorithms simply can’t replicate that effectively yet (even the very few who’ve managed to navigate the Turing Test!).
- Jobs in which you can increase your own value. Nearly all positions have a set of base level performance indicators, but many offer considerable scope to go beyond the bare minimum and take on extra responsibilities or expand your field of influence. Over time, the remit of the job, therefore, evolves with you, and the niche you end up filling for your employer effectively becomes a bespoke one. Machines can’t really do this: they have a single role with strictly defined parameters, and will never do any more or less than explicitly instructed. Not all bosses would necessarily see that as the definition of a great employee.
- Jobs that frequently raise new challenges. If each work day serves you up a fresh set of obstacles to overcome (the sort that can’t necessarily be tackled using a one-size-fits-all approach, even if they share a basic theme), then the chances are an algorithm might struggle to perform well in the role. Machine learning can certainly make a judgment call based on precedent, but it offers very little in the way of flexibility or adaptability.
- Creative jobs. Another obvious one, perhaps, but we’re still a long way from seeing the first AI author scoop a major poetry award, move us to tears with a beautifully delivered monologue, or – perhaps more relevant to most of us – even come up with a reasonably compelling three lines of marketing copy. And, despite having developed robotic chefs that can cook incredibly precise meals by following a downloaded recipe, their attempts at creating recipes from scratch have been a bit…well, like this. Delicious, right?
