AI: the future of work (hallelujah)!

A lightly edited version of a talk given at St. James Clerkenwell, June 26th 2019

Tonight I’d like to ask two questions:

  1. Is a Robot going to take your job? 
  2. And what has faith got to do with the future of work?

Screenshot 2019-07-01 at 10.31.29

First of all, what is AI?

AI is computer software that is designed to mimic some of the functions of human intelligence. Right now, it mostly functions like a calculator on steroids. Good at analysis,  bad at common sense. AI mostly works for narrow, bounded, analytical problems when you have the data to tell the algorithm what the right answer is for hundreds or thousands of different cases.

For example, games. These are generally a known combination of rules and moves toward a specified goal. This is a field in which the analytical intelligence of AI can excel. Indeed, for the first time in 200,000 years, there now isn’t a board game that computers aren’t better at than people. Within 30 years, I suspect, there won’t be many games, physical or mental where AI algorithms can’t beat people.

And that’s ok! Computers have been beating people at chess for two decades now, and more people play chess than ever. Wikipedia doesn’t stop my enjoyment of pub quizzes. And just because there are robots that can run faster than I can, doesn’t spoil my enjoyment of the Olympic games.

While the individual algorithms are very specialist – the applications of AI as a technology is very general. AI is so good at calculation, that it can take relatively unstructured information, discover complex patterns within that information, and make calculations on those patterns to ‘learn’ what the right answer looks like. That’s why, unlike your pocket calculator, AI is able to take images from a camera, and radar, and other sensors, and learn how to drive a car by recognising images from their component pattern of shapes. Or accurately identify terrorist propaganda on the internet with superhuman speed.  AI has also been able to reach human levels of performance in voice transcription, translation, and even more ‘creative’ tasks like painting and musical composition.

But while displaying superhuman levels of speed when it comes to analysis Artificial intelligence is really bad at other tasks that wouldn’t tax my young son. Like the answer to the question – “what’s going on”? AI works with quantitative data, and fast feedback, but don’t deal well with context, or anecdote, or personal relationship. AI is a peculiarly narrow, mechanical form of intelligence.

Screenshot 2019-07-01 at 10.37.12

How is AI going to change our work?

Well, there’s good news and bad news.

The good news is that tasks that are repetitious, predictable, and analytical, can to a large extent be automated. I’m thinking of working out how to set prices in a shop, or working out the maintenance schedule for a fleet of trains, identifying who might like to make a charitable donation, or constructing a staffing roster for a hospital (other examples here). This is usually a piece of someone’s job, but it is often a pain the neck. AI will help people to do less brute calculation, and have more time for more interesting, more human work.

More generally – there is enormous promise in the potential for AI to benefit society. In aggregate, this will amount to higher productivity, higher wages, and lower consumer prices for stuff. But perhaps that sounds rather dry. Likely specific applications of AI are more colourful – for example: 

    1. Making personalised tuition affordable for the many not just the few;
    2. Being able to diagnose our health faster and more easily;
    3. AI in self-driving cars should reduce city congestion & the price of houses within commuting distance of work;
    4. And though our energy markets, AI could even help to tackle climate change

The bad news is that some jobs are mostly repetitious, predictable, and analytical. Like taxi-driving, call centre staffing, basic accounting audits and compliance functions, paralegal work. AI will replace jobs that centre on basic cognitive skills, just as machines have replaced much manual labour. These jobs are likely to go, or at least change radically. 

Screenshot 2019-06-24 at 11.06.22

But not all at once, and not quickly. AI will diffuse through the economy gradually over our working lifetimes. Just like other general purpose technologies have through history, like the containerisation of shipping, or the mechanisation of agriculture, or the electrification of factories. These changes take a generation to really reach a critical mass. More often, people and AI will work together, just as people use machines to support their physical labour now. Over the last century these technologies have been the driving force behind rising wages and the standard of living for everyone, and I believe the same will be true for AI. Looking back – there are few technologies that we would wish to uninvent – other than perhaps nuclear weapons. And the selfie-stick.

But the impact will be less predictable than we might assume from the hungry headlines that suggest robots are going to take all our jobs. Indeed – there have been headlines saying that for the last 200 years! Take bank tellers – after the introduction of ATMs, they weren’t replaced – their numbers rose as the ATMs made the banks more efficient and allowed the people to focus on more complex tasks like customer support, or loan analysis. Today you might hear that we shouldn’t be hiring any more radiologists; that AI powered diagnostics are going to do all the work. But I suspect that AI will actually allow nurses and doctors to focus on communication and empathy with the patient in a way that both they, and patients currently miss, when appointments might last just 7 minutes and much of that time is spent typing into a computer.

ai-unemployment.png
Source: The Economist

We have always been pretty bad at predicting what jobs will come next. There are many fewer washers and agricultural labourers than there were 100 years ago, and people were similarly anxious. But there are many more accountants, and developers, and bar-staff. My grandmother could hardly have predicted the jobs that many of you are currently doing today. I think you will be just as hard pressed to forecast the working lives of your own grandchildren.

Lest you think me just another tech fanboy, it is worth putting my general optimism in the context of some different, and subtler risks that AI presents. We don’t hear as much about these, but we should be actively preparing for them:

  1. Opacity that disguises selfishness – lowering costs, while also lowering performance standards.
  2. Complexity that creates fragility – rather like modern supply chain management. This can create a hidden fragility that may cause nasty surprises.
  3. Technology that pretends objectivity → people subjectively create algorithms. They can still be just as biased and discriminatory as people can be.

And finally, Labour market change hasn’t always been handled well – the deindustrialisation of Wales and the North of England in the last quarter of the 20th Century left a scar that has ached for a generation.

But overall I am optimistic. Across the country, more people change jobs every year, than will be displaced by AI over the next ten years. And there are plenty of things we can do personally, and as a society to support the introduction of tech in a way that minimises the impact of change for people.

What should we do to mitigate this transition?

There are some things that governments should consider through public policy – like funding retraining schemes for available local jobs, providing wage insurance to affected sectors and professions, and extending schemes like the Earned Income Tax Credit. But I mainly want to focus on the things that we can do ourselves, and in our communities.

We should keep learning. Few of us ever learn at the same rate than when we are at school or university. But there is no reason for that to be the case, other than culture. We have been talking about life-long learning for ages, and the sum of human knowledge is accessible on the supercomputers in our pockets. But for a bunch of reasons it isn’t yet simple enough for people to pick up new skills. When it comes to AI, one of the reasons it induces anxiety is that we feel like change is being done to us – that we’re not in control. Why not then take back control, and actively resolve how you would like to apply AI to your own work? And for our children – perhaps consider changing the standard question – “what they want to be when they grow up” to “what five things would you like to do when you grow up?”

What we learn also needs to change. The education system is understandably conservative, but in a world where access to knowledge is abundant we should perhaps move away from the memorisation of facts, towards a more explicit orientation towards open skills that are applicable to many jobs – like analysis & critical thinking, history & politics, communication & storytelling, as well as the building blocks of learning like Maths and English. None of these skills are going to be automated in our lifetimes. Moreover, we’ve long fostered conscientiousness within school as a trait that suits us for working life, but I suspect adaptability may become just as important if not more so, despite being anti-correlated.

Most importantly to help a transition to a world of AI, I think we should be clearer about the differences between machine intelligence and human intelligence.

There was once a sign on the wall of the NASA control room that said – “In God we trust; all others bring data.” But people have more to bring than just data, and we need to make room for that at work. In 1986, a NASA engineer tried to get a shuttle launch stopped, because of a concern he had about the safety of some components. But he didn’t have the data. And the launch went ahead. The Challenger space rocket exploded shortly after take-off.

nasa

It is a cliche of management textbooks, that if you can’t measure it, then you can’t manage it. Is that how our marriages work? Is that how any personal relationship works? Working with AI shouldn’t involve an empirical reductionism, but should involve us bringing those facets of human intelligence – relationships, and anecdotes, and history, and emotion, and meaning – together with calculation and computation. That is what makes us human, and what makes up 9/10ths of the world’s wonder. AI will tend to be designed to optimise efficiency, but that efficiency should never come at the expense of significance.

What’s Faith got to do with this?

As well as working in the field of AI, I’m also a Christian. I try to fit my work to my faith, rather than just accommodate my faith to my work. But the Church has long had an uneasy relationship with technology. The current vogue for tech pessimism from academics like Shoshana Zuboff has its roots in the work of Christian philosophers like Jacques Ellul who have long cautioned about confusing efficiency with purpose. From this general rejection of technological utopianism, the Church has made predictions for at least ten out of the last two apocalypses. A marginally better record than the number of recessions forecast by mainstream economists, but still not a reliable guide to the future.

I don’t believe that the arrival of AI signals an impending apocalypse, but I do think faith has plenty to contribute to the question about how we might use this new technology well. Many of our public and private conversations about faith tend to focus on whether it is true. Perhaps a less discussed but equally important preceding question is whether faith is useful. I believe Christianity is particularly useful when thinking about the future of work. Specifically I think there are three ways in which faith has influenced my own approach to work:

  1. First, faith helps us towards vocations over careers. Vocations tend to be worked out in service to others or to God, and pursued with a clear sense of mission. Careers tend to focus on our own glory and reward. Careers follow the enticing logic of individualism that permeates so much of contemporary society, whereas vocation and service are the pursuits that tend to generate lasting fulfilment and joy.
  2. Second, faith helps us to get beyond the measure of our lives being whether we are ‘making a difference’ or ‘changing the world’. I’m all in favour of changing the world, but faith reminds me that I am a human being, before I am a human doing. All people are made in the image of God, whether they are employed, unemployed, or unemployable. Our inner lives, and our relationship with God, matters even more than our faint scratches on the consciousness of the world.
  3. And third, faith tends to afford a deep sense of peace and confidence in the circumstances we find ourselves in, despite our own human limitations and fallibility. For me, that the source of the trouble in the world isn’t principally artificial intelligence, or Brexit, or rising inequality – important as those things are – but in the division of our own hearts. Faith helps us to internalise that while we can’t always have what we want – more money, more friendship, more time – we can almost always get what we need. As the character Sonny said in the film The Best Exotic Marigold Hotel, “Everything will be alright in the end. If it’s not alright, it’s not the end.

Now these three aren’t unique to the faithful. There are people who don’t believe in God who have thought about life from first principles and often come to the same conclusions. But I’ve certainly found Faith is a hugely helpful mental scaffolding to navigate the challenges of working life – especially when faced with unsettling change.

Conclusion

So to sum up – is a robot going to take your job? Probably not. AI is more likely to be a blessing than a curse – but over the next couple of decades we will notice its impact more and more. There are things we should do – re-train, re-skill, reconsider how human intelligence sets us apart from machines, and help people around us to the same. But in addition, the lens of faith might help us to resolve whether we are working in the best way – vocation before career, being before doing, and confidence before anxiety.

That’s my view, but I’d love to hear yours. Before the Q&A we might want to stretch our legs, and to introduce ourselves to the person sitting next to us. Why don’t you try and find out which aspects of your neighbours job they would most like to automate using AI?