Is everything in the world the result of emergence? The natural world is the product of evolution, an emergent process: accidental mutations and impromptu encounters between one organism and another, or an organism and their environment. In this sense, the whole of human civilisation can be regarded as emergent – including even the most dictatorial leaders, who appear imposing but emerge out of groundswells of populism. Karl Marx described Das Kapital as a study of natural history.

But within the broad reality that everything bubbles up from below, there are some phenomena in nature and society that look very top-down. As anyone who has watched David Attenborough documentaries will recognise, lions and whales display hierarchical behaviours; and to jump to the example of British politics, power that was once more distributed among free-thinking backbench MPs and relatively autonomous local councils is now highly concentrated in the hands of the prime minister.

Sorting the world into top-down and bottom-up can be illuminating. Such binary oppositions feel especially apt in a world ever more divided between the tech titans and those that deliver food on electric bikes. But the model has limitations. It’s sometimes hard to say where the bottom-up ends and the top-down begins, and what looks bottom-up can contain elements of top-down on closer inspection – and vice-versa. The interplay between happenstance and strategy is everywhere.

If you put different kinds of human cells onto a petri dish, Jonathan Chubb told me, they will grow into organoids; rudimentary versions of the organs they were destined to be. Brain cells will grow into a basic brain, and lung cells into a sort-of lung. ‘There’s an innate tendency for different cells to organise themselves spontaneously in space without anyone telling them what to do’, Chubb said; but to grow a real organ ‘you need some kind of coordinate system or top-down regulation’. Top/bottom binaries are ‘useful concepts’, he said, but ultimately ‘the full richness of structure that you get in an animal is always going to be a blend of both top-down and bottom-up.’ This is true even of slime moulds: ‘we tend to think of Dictyostelium being the most emergent of all systems’, but even here the building of spores involves some ‘top-down information being actively provided’.

David Krakauer, president of the Santa Fe Institute, a multidisciplinary university in New Mexico dedicated to the study of complexity, agrees. ‘I don’t like bottom-up and top-down language’, he said, ‘I prefer local versus global.’ He points to the example of a brain that is connected to every part of your body, or, for that matter, a representative democracy: both behave more like a network, or a ‘hub and spoke graph’. In a democracy, the elected leader is the hub and citizens are the spokes; and in any case it’s ‘more complicated’ than top versus bottom because ‘we build institutions where we delegate power for a period of time’, or in the case of biology, ‘all the cells in your body have agreed that it’s useful to have a brain’: they can say ‘I’m happy for you to decide where we go for dinner’.

Although imperfect, I find I can’t let go of the bottom-up versus top-down binary, because it provides a way to articulate how monopolies make false claims to grassroots legitimacy. Big tech is a prime offender here, with insiders often describing the internet as emergent, and even comparing it to slime mould (Physarum has been called a ‘natural computer’). They are influenced in part by Stewart Brand and his Whole Earth Catalog, which envisioned the early internet as a countercultural egalitarian network emerging as a reaction against hierarchy and authoritarianism. Analogies between slime moulds and AI, including large language models (LLMs), abound. The suggestion that these technologies have evolved naturally distracts us from both the highly top-down ways they’ve been engineered and the benefits they promise to already wealthy interests.

A further complication, however, is that in certain specific senses computers do learn and operate in a bottom-up way. When neural networks were first imagined in the 1940s and 50s, they were ‘more like ants or cells’, David Krakauer told me: researchers connected up computers as if they were nerve cells in the human brain and asked ‘what could they do if I didn’t give them very much at all’, but just ‘wired them together’. The first robots were trained using the paradigm of top-down human intelligence. But then companies like DeepMind had a revelation: a much better way is to build them bottom up, by modelling them on infants, those amateur empiricists who learn through thousands of micro interactions with their environment; for example, in my case, seeing what happens when you stick a frozen pea up your nose (answer: tweezers). Researchers are using slime mould to design new kinds of computers, such as rather alarming-sounding ‘living devices’. The very structure of recommendation sites – if you like that, you’ll love this – is itself a reinforcing pattern, although arguably an example not of emergent intelligence, but stupidity.

The concept of emergence, and analogies between emergence and artificial intelligence, arose out of long-standing debates about the mind-body problem and the question of consciousness; how ideas emerge from meat. ‘There was no theory we had that could explain how a physical mechanism would produce conscious states of mind’, Krakauer said. But then ‘LLMs, come along’, and people say ‘Yaha! We’ve made them now, and it turns out they’re like plumbing, too’. What we seem to be ‘captivated’ by, he continues, is the fact that ‘you could just throw all those stupid or rather ignorant bits together and they do something that’s a bit astounding’; and that’s why the ‘tech bro science militia’ is ‘feeling smug’. But instead of making a physical computer into an intelligent mind, he says, they’ve ended up doing the reverse: ‘they’ve basically turned the mind into a body’.

Comparing AI to largely emergent slime mould is ultimately ‘misleading’, Krakauer said, because ‘it’s just not bottom-up at all’ in the sense that ‘huge amounts of human knowledge, more libraries than we could imagine, have been consolidated into training these things’. In fact, LLMs are ‘the opposite of the slime mould story’, he says, and ‘it’s peculiar that we’ve been brainwashed into thinking they’re intelligent’, when LLMs are in fact not intelligent, but simply ‘knowledgeable’.

‘When I was at school’, Krakauer continued, ‘there were people who would just effortlessly breeze through problems, and they never did their homework’. This was ‘very annoying, while the rest of us were busy studying’. He suspects that ‘the emergent argument rhymes with this idea’ and concludes: ‘I think that’s why we like slime moulds. They seem so stupid, and yet they can still be so amazing.’ Slime moulds are the orchestra without a conductor that miraculously manages to perform a symphony.

Portraying AI as emergent is not the only way in which the semblance of the bottom-up is instrumentalised. That ‘tech bro science militia’ holds out the promise of YouTube stardom to amateur content creators; political astroturf campaigns create the illusion of mass support; and mega-brands hide behind quirky fake-unique café facades. Emergence is an enlightening paradigm, but it can also be deployed as cynical PR.

Not only that, but I believe we should continue to question why emergence is so appealing. In an atomised and post-ideological era, the idea of collective actions producing outcomes that are more than the sum of their parts shimmers as a chimera of bygone ideals. In their absence, we cling to the belief that it is possible for ordinary people to achieve their dreams, and enthusiastically grab at instances of bottom-up exceptionalism, turning a blind eye to the orchestration behind the scenes. Wikipedia is in fact updated by a relatively small number of tech-savvy volunteers; in commentary about the Arab Spring and Citizens’ Assemblies, the role of self-selecting and unusually well-informed activists tends to be similarly underplayed.

Self-organisation is a well-publicised factor in the development of computer games, facial recognition systems and voice-recognition software: all technologies that, paradoxically, are also characterised by top-down social control. In his 2002 book on emergence, the science writer Steven Johnson speculated hopefully about ‘the future of artificial emergence’ and ‘what will happen when our media experiences and political movements are largely shaped by bottom-up forces, and not top-down ones’. Such optimism feels misplaced in 2025.

The claims that are made for emergence serve to obscure the increasing dominance of powerful interests, therefore, but also the underappreciated value of authority. What is praised for being spontaneous could sometimes do with more coordination: the Occupy movement, for example, largely failed to make lasting political headway because it eschewed the very leadership that could have articulated and taken forward key demands. Unbound, the crowdfunding publisher, went into administration this year, exposing as a sham the notion that we don’t need the traditional publishing industry, with its stuffy old standards. ‘I thought, “someone has to build a new kind of publisher, one where an author doesn’t have to rely on a gatekeeper to decide whether a book is commercial or not”’, Unbound’s founder, Dan Kieran, told The Bookseller in 2014. It turns out that gatekeepers have their uses after all.

In The Wonders of Life, the physicist Brian Cox urges us to admire the ‘humble majesty of a blade of grass’ and notes that it shares much of the same biochemistry as us humans. ‘I suppose this is a most difficult thing to accept’, he writes, since ‘the human condition seems special; our conscious experience feels totally divorced from the mechanistic world of atoms and forces, and perhaps even from the “lower forms” of life.’ Yet this feeling, he believes, is ‘an illusion’.

We should marvel at what an unprepossessing creature like slime mould can do. But let us not fetishise either lower forms of life or ChatGPT, because the tech bros are only too happy to downgrade our cerebral capabilities in favour of the machines that one day may kill us all. In a culture that glorifies the hive mind and disparages intellectuals as out of touch elites, it’s easy to lose sight of the fact that sophisticated ideas, professional judgement and effective chains of command can be good for us. Our brains work like slime mould in the sense that both are emergent; but both also, to a crucially varying degree, combine the bottom-up with the top-down. We undervalue human intelligence at our peril.