How can AI augment rather than dictate human action? This expert explains

Can machines imitate humans? Alan Turing, the famous British mathematician and computer scientist posed this question in 1950, devising an experiment to explore the idea of how and whether machines can think. It’s a question that almost 70 years later still fascinates humankind, and one that has become inextricably linked with perhaps the most well-known of the emerging technologies: artificial intelligence (AI).

AI, and the need to marshal it as a tool and a resource, was the focus of a recent session at the Annual Meeting of the Global Future Councils (AMGFC) in Dubai, given by Professor Erik Brynjolfsson, at the start of which, he suggested that technology – specifically the invention of the steam engine – is the most important thing that has happened to humankind.

Until the Industrial Revolution of the 19th century, the average person lived close to subsistence level, as had every generation before them. Once the technology to harness turning water into steam, and with this power, was invented, GDP per capita skyrocketed. Today, we enjoy 50 times greater GDP per capita than our pre-Industrial Revolution forebearers.

Brynjolfsson reasoned that AI offers the same transformational opportunity because it exhibits the same three characteristics as the steam engine: it’s pervasive, can be improved over time, and will spawn complementary innovations. In essence, it will drive an arc of improvement.

We’ve now entered an era where AI is – like water being transformed into power – crossing a threshold. It’s doing ever-more of the tasks almost as well – and in some cases, better – than humans. AI, it seems, can imitate humans.

This is creating a maelstrom of challenges and opportunities, as well as raising a plethora of questions. A basic, if fundamental starting point is that if AI can do the same tasks as humans, does this spell the end of work? Brynjolfsson suggested that AI will transform work, but for him, it is a case that there is no single job that AI will take over, but nor will there be any job where it has no effect because it’s the task not the job that becomes automated.

Taking the example of a radiologist, Brynjolfsson explained that this role has been broken down into 27 tasks, some of which – such as diagnosis – AI is proving better able to do, but others – such as administering sedation – where a human is by far the preferred option.

This is just one sort of role, and what Brynjolfsson suggested is the need for a systematic approach to addressing how best to both manage the transformation of jobs and work, and realize the opportunity – in terms of gains in areas like productivity and efficiency – of AI’s part in this.

He argued the need for senior managers and policy-makers to implement a task-based analysis, pursuing a dual approach: to first develop a plan that identifies and ascribes a value to every task of every job role, and secondly, to identify opportunities and track progress using clear KPIs that don’t rely on old metrics. Instead, indicators like customer satisfaction and employee turnover will help reveal the areas where AI can support real business and economic value.

He argued that this will also shorten the time from the invention of the technology to delivering better productivity, in turn boosting GDP, and the hitherto unmeasured environmental and health benefits.

Returning to Turing’s Test, these days we might ask whether we want machines to imitate or think like humans? For Brynjolfsson, a better line of enquiry is to see where AI can augment what humans do, using this technology as a tool to make a job better – whether more efficiently, to a higher quality, or more creatively – rather than replace it.

Transformation is a modern buzzword, and for Brynjolfsson it typifies the coming five to ten years. This is not just any transformation, but the largest humankind has ever witnessed. Initially, Brynjolfsson argued, we will see a dip in growth.

He likened this to a J-curve, in which there’s a productivity slowdown as companies develop new products, reskill their workforce, and create the intangible assets that can’t be measured but support the take-off of growth once they’re in place. This period can be discouraging, but it’s been a familiar pattern during the introduction of other technologies, and one that Brynjolfsson argued is already evident with AI.

He also suggested that we face a dangerous decade because in terms of organisational structure, institutions, legislation and economic understanding, we aren’t gaining a sufficiently rapid understanding of the changing dynamic.

What we want to avoid is the use of technologies like AI to centralize decision-making and power. Instead, what’s required is greater investment into areas like economics, law and sociology to better understand the wider policy implications of this technological transformation. Only if the right decisions are made, do we stand to realize the promise of the large-scale benefits for the planet and humankind that technology offers.

For Brynjolfsson there is no single, inevitable future. Instead, how our future is shaped depends on the choices we make now. AI is a powerful tool, offering a tantalising opportunity to make sizeable changes to the world. To make the right choices, Brynjolfsso argued, we must focus on our values and goals, and steer AI’s development to meet these.

Ultimately, what we don’t want is what Brynjolfsso described as the Turing Trap – being in the position of having machines perfectly imitate what we do. Instead, as humans we should play to our strengths, and use AI to augment our world, not dictate it.