The Atomic Human: Understanding Ourselves in the Age of AI, Neil D. Lawrence, Allen Lane, £25

In 2009 Mark Zuckerberg delivered his prime directive to the staff at Facebook – ‘Move fast and break things.’ In 2014 he relented: ‘Move fast with stable infrastructure.’

Zuckerberg had discovered that, thanks to his growth plan for Facebook, he had relinquished control of his creation. Facebook was defenceless against bad actors. The company had no idea what was going on as political lies and mass information trawling turned the site into a swamp. The Cambridge Analytica scandal, in which the company scooped up the personal information of 87 million Facebook users, sucked the fun out of breaking stuff. The machine had gone rogue. Facebook still doesn’t know how its overall system operates, and it still doesn’t know if it is being manipulated by those bad actors.

They don’t care, they just need our info. As Neil Lawrence powerfully observes: ‘The emergent digital oligarchy derives its power from the aggregation of our personal data.’ But, in the first years of Silicon Valley’s multi-billion dollar tech frenzy, breaking things seemed entirely legitimate. Computers’ storage and speed accelerated to the point where, it seemed, they could do anything. Data on Star Trek, Terminator in, er, Terminator, Sonny in I, Robot and countless others had filled the popular imagination with the idea that we could, very soon, construct humanoid machines. The message of this extraordinary book is: we can do no such thing, nor should we try. We can, however, build artificial intelligences that may well fill the gaps in human competence.

Neil Lawrence is Professor of Machine Learning at Cambridge. He has, among other things, worked for Amazon. His book is a hybrid of autobiography, science history, industrial history, geology, biology, Renaissance art, multiple biographies and, well, you name it. He spends time on the way the Allies won D-Day and Montgomery beat Rommel in the desert on the back of information-gathering and the genius of, among others, Alan Turing. This range leads to a fanatical degree of digression, which, for a while, irritated me. At one point I am just grasping the significance of T-cells and cytokine storms when I find myself at the Battle of Kursk in 1943. But stick with him, it all makes sense in the end.

The title refers to the ultimate point at which we can always distance ourselves from the machines. ‘As the machine slices away portions of human capability… if we are left with something,’ he writes, ‘then that uncuttable piece, a form of atomic human, would tell us something about our human spirit.’

His faith in that uncuttable piece is paradoxical. It is based on human limitation.

‘Machine intelligences,’ he writes, ‘are defined by their capabilities, but atomic human is defined by our limitations.’

We all suffer from a form of locked-in syndrome. When two computers meet, they, all at once, know everything about each other. Everything I write here is known at once to the Apple Cloud, to my laptop, and my phone. When two humans meet, each knows almost nothing about the other, even if they are old friends. We have to feel our way through talk and observation, but, however good we are at that, we will never attain anything like full knowledge. The mystery of the human other is the story of our evolution and our lives.

Add to that the 65 million years since the first primates or the mere 1.3 million years since the first hominids left Africa and it becomes clear that time has been our teacher. The more we strived to communicate, the more complex our minds became.

Our artificial competitors – the first digital computers – appeared in the 1940s. The more we used them, the more we came to believe in their power. In truth, the computer on your desk is no better than a screaming baby, however much it may try to convince you otherwise.

Yet Lawrence does talk about ‘machine awareness’. He detects it in the ‘centrifugal governor’ developed by James Watt in 1788. This allowed a steam engine to regulate the flow of steam without human intervention – ‘…the machine gained awareness of its own state’. This may seem a bit of a stretch but it is, as Lawrence says, ‘a large conceptual leap’. We are now surrounded by machines aware enough to run themselves, even to repair themselves. It is all too easy for excitable or pessimistic minds (or just very rich people) to assume the machines will inevitably win and we will be dominated by ‘new cognitive monsters’.

And they might win. They have, Lawrence, attained a ‘new level of cognition that we can think of as System Zero’. This refers to the way computers can now second-guess us and prejudge what we want. Simple things, such as recommendations on Netflix or Amazon, are based on data-trawling on a scale so vast we can barely imagine. Such devices restrict our view of the world but, at the same time, they give us a feeling of being known by the machine.

‘By giving away the key to our digital soul, by giving away our personal data, we give up our personal freedom.’

The problem is that System Zero ‘doesn’t understand social context, doesn’t understand prejudice, doesn’t have a sense of a larger human objective, doesn’t empathize’.

The reality is, in Lawrence’s words, ‘machines only work because we take away uncertainty’. The ranks of humming machines that consume your information are coddled in special rooms; unleash them into the outside world and they would be helpless. We, in contrast, live in a raging ocean of uncertainty which we can control with the ingenuity of our locked-in mind. The machine is, again, babyish, fed and stroked every second of their lives.

But – and it is a big but – they can ‘communicate without our restraints and so it is possible they could come to dominate our lives’. The book concludes with a warning that ‘we must step up’ and ensure the machines remain our servants. He is right – of course we must, but will we? There is, as we should all have learned, no limit to human folly. The faces of the gamers, the coders, the geeks, the freaks and Mark Zuckerberg, when they are absorbed by their machines or when they talk about them, suggest a quasi-religious intensity. But the machine god would be no friend of ours and it certainly won’t have read or understood this astonishing book. So don’t break things, you might need them later.

In conversation with EI’s Paul Lay and Alastair Benn, Neil D. Lawrence challenges received wisdom on our AI future. Listen in through the link below.