The possibility that AI will outsmart, and eventually outlive, humanity is real, if distant, and it is right to be thinking hard about it now. But media hooha over a speculative future should not deflect attention from the changes already surging through the world of work resulting from the introduction of the burgeoning array of first-generation generative AI apps.

BT recently announced that AI would soon replace 10,000 of its customer service jobs, while adland’s Sir Martin Sorrell went one further by calling time on the call centre, full stop. Wall Street is said to be using AI to rewire the business of finance, hiring as well as firing. The CEO of high-riding chipmaker Nvidia, Jensen Huang, sees the arrival of the new apps as the ‘tipping point of a new computer era… Everyone is a coder now. You just have to say something to the computer.’

Let’s step back a bit. 

Like any other technology, of itself AI is neither good nor bad. It can be used for both. But AI is also not like any other technology, for at least two reasons. One is its combination of potency and breakneck speed of uptake, the fastest  ever – and currently far outpacing the speed of regulatory thought. And the second is the juncture at which it arrives, marked by a technological and economic determinism that threatens to make the outcome overdetermined.

At its simplest, AI presents us with a choice. In the deceptively simple words of the distinguished MIT Prof Erik Brynjolfsson, a signatory of the most recent open statement of alarm over the direction AI is taking, will it be used to augment or automate human labour? 

For there is a vast and much overlooked difference between the two. While automation is often billed as the most important technological advance since the industrial era, it is actually augmentation that delivered the major productivity and market-creation gains of the past. A moment’s thought tells you why: humans plus technology are capable of vastly more than humans on their own. Think skyscrapers, modern medicine, air and space travel, the Toyota Production System. In combination they create new industries and markets, and crucially the benefits have traditionally been shared through jobs. A crane operator is more productive, more skilled and better paid than a bloke with a wheelbarrow, a radiographer than a barefoot doctor, a jet pilot than a carrier with a horse and cart.

But the reverse side of this coin is that, also counterintuitively, pursuing human-like AI as Holy Grail, as researchers and technologists have done and still do, is a trap. While ‘augmenting humans with technology opens an endless frontier of new abilities and opportunities’ (Brynjolfsson), the more AI mimics the human, the greater the temptation to use automation to carry out the tasks that people already do, only faster and (at best) more cheaply. 

That’s what is happening now, and it is doubly perverse. Getting technology to do things that humans have evolved to do easily – how long since we were promised driverless cars? – is both harder and less valuable than making it do the tough and boring stuff, such as instant complex calculations or holding unlimited numbers of things in memory at the same time, that deflect humans from using their ingenuity to make the counterintuitive leaps that trigger the most productive innovation.

Spare no tears for the call centre, a previous misbegotten attempt to use technology to solve customer problems that are better dealt with by people. But if BT simply automates unsatisfactory existing processes without rethinking them first – the classic digitisation investment mistake – it will compound the error, neither improving dismal customer service nor saving cost. 

This kind of automation creates little or no value for anyone except execs and shareholders. For employees, whose value resided in knowledge that is now codified and turned into capital, it is entirely negative. Not only that, the advent of AI, with its growing ability to monitor and micromanage work, effectively kills off the idea of the job as the primary distribution mechanism of the benefits of technology at work, throwing a whole system into crisis. An increasing number of companies no longer have any interest in humans as labour, only employing them as a last resort.  

This gets to the heart of the matter. Ironically, the driver of this kind of valueless automation – think self-service and contactless supermarket checkouts, digital-only services of all kinds – is the ultimate example of Brynjolfsson’s human-plus-technology augmentation: what we now call surveillance capitalism. The unholy alliance of internet, algorithms and advertising, surveillance capitalism deploys human-focused technology to bolster corporate and shareholder interests through tracking and targeted advertising. In the surveillance business model that has now captured almost all companies, humans remain their greatest asset: but as the source of raw material to be exploited and sold as information to others, not as workers.

Prevailing economic discourse around AI displays a kind of studied fatalism. ‘Of course, advancing technology means jobs will go. But in the long term it will create new types of better work that haven’t been imagined yet, as it has done in the past. If not, then government will need to provide.’ This is pure bad faith. Technology isn’t destiny. As Daron Acemoglu and Simon Johnson, as MIT professors hardly technophobes, spell out in their new book Power and Progress, there is nothing automatic about such outcomes. How technology plays out in the workplace has always depended on choices made in boardrooms and finance departments (or sometimes throne-rooms or lordly mansions), which in turn reflect power, material incentives and evolving social norms as embodied in legislation and regulation. Positive outcomes, they make clear, have to be fought for.

Realistically, it’s clear that AI, or machine learning as it is useful to call it from time to time, is a general purpose technology that will end up everywhere. There are dozens of apps available already, with multiple new ones coming to an app store near you every day. AI will eat many jobs and create some new ones, although probably fewer than we would wish.

Equally clear is that without intervention AI will do little to counter the major social problems we are currently struggling with – particularly the inequalities that fuel discontent, social and extreme populist politics – since it comes from the exact same parentage. Bryjolfsson argues that business, technologists and policymakers have all put too much emphasis on automating rather than augmenting labour, and the first imperative is to eliminate or reverse the tax and other incentives that favour it. Acemoglu and Johnson propose a range of other measures, ranging from breaking up Big Tech and removing its immunity to prosecution for the content it carries to boosting countervailing forces, abolishing patent proection for surveillance technologies and imposing a digital advertising tax.

If some of this had been enacted the first time round, particularly measures affecting advertising and surveillance, the new AI apps wouldn’t seem such a frightening prospect. But for once, wiser from experience, we have a second chance. Ths time we had better take it.