Let’s Be Inefficient!
Tech, Culture and the Risks of Optimising Everything
Recently, two bigish ideas have kept popping up in my head: culture and efficiency. And of course in the context of frontier technology, especially artificial intelligence.
Culture, because its all around us and makes us who we we are. Efficiency, because that’s what big tech constantly (and aggressively) peddles.
It dawned on me that we don’t easily relate each of these to the other. It seems like they really don’t necessarily go hand-in-hand. (Think about the last time you said to yourself: “Geez, why do they do it THAT way? What’s the logic?”) Culture isn’t always efficient and its not always meant to be so. But I’m afraid that Northern tech is slowly forcing an efficiency culture on us.
Efficiency demands minimising waste, streamlining processes, and maximising output (and profit). Culture, on the other hand, is inherently messy, layered and (let’s be honest) not necessarily performance driven! And therein lies both what gives its strength and makes it vulnerable. With big tech obsessed with optimisation, anything that resists streamlining becomes a problem to be solved or eliminated.
Remember not too long ago when your taxi driver (or you!) had to stop to ask for directions? The milkman? Buying books in the local bookshop? Or in Thailand today, the women delivering that probiotic drink Yakult on their motorbikes (average cost is around THB 18 per unit)? All these encounters create serendipitous moments, opportunities to exchange gossip and news, and to acknowledge our shared humanity. All these exchanges, on the surface, aren’t terribly efficient.
But one of my fears is that in the race to build AI systems that optimise everything, we risk erasing those very “inefficiencies” that bind us together.
Let me unpack this a bit
Culture has been top of mind for me ever since I wrote The Moral Lighthouse [1] and through the work I did recently with Malaysia’s National AI Office (NAIO) around cultural values in AI governance and ethics (work I I did with my fabulous new colleagues and friends Dr. Eric J.W. Orlowski and Prof. Chithra Latha Ramalingam). Then I read a piece this week from Eric and his colleagues at AI Singapore (Hakim Norhashim and Tristan Koh Ly Wey) about cultural alignment in training models for AI systems (entitled “Taming Cultural Fractals: Why AI Needs to get ‘Thick’”) which got me thinking even more. [2]
Their piece nods to anthropologist Clifford Geertz who emphasized an interpretive theory of culture, understood as a system of shared meanings and symbols that people use to make sense of their world and guide their actions. He called this the “thick description,” moving beyond surface-level observations to delve into the context and interpretations behind cultural practices. In other words, the web of meaning around us and in our activities – context, emotion, memory, belief. As Eric explained to me in a WhatsApp exchange,
“I suspect this is why developers are struggling with culture and AI models. Because it can’t be meaningfully reduced to statistical components.”
But “efficiency” — that’s a word that strikes at the heart of neoliberal capitalism. It encapsulates the relentless pursuit of profit (often framed as “shareholder value”) through the ongoing optimisation of marginal productivity, market performance and economic growth. Anything deemed “wasteful” (another version of inefficient) must be eliminated and every dollar of input must yield maximum return — for the first as well as the last dollar. Inefficient friction is a problem that must be engineered away.
Take the logic far enough, and you arrive at the usual neoliberal prescriptions: deregulation, privatisation, and minimal government intervention. (Of course, we also know where this path has led us — widening inequality, shrinking social protection, neocolonialism and the steady erosion of the welfare state. But that’s a story for another day.)
In tech, some inefficiencies are removed to “improve experience” while others (oddly) are intentionally designed to frustrate users, especially when nudging them not to act against a company’s interest so that users drop off or give up on a particular flow. (Think about the last time you tried to deselect various kinds of cookies, including those pesky “vendor preferences” that you know, like me, you’ve given up on halfway through.)
So in the neoliberal sense of Silicon Valley (and its Washington vanguard), efficiency is considered a moral good. When could being inefficient ever be a desirable final or even intermediate outcome? For engineers and tech executives hawking profit-making tools, this can’t possibly lead to an optimal outcome – eliminating friction shows progress! Efficiency or bust!
But this obsession with optimisation often hides deeper costs, especially when applied to AI systems that will interfere, and shape, human systems.
The Sales Pitch
As big tech continues to develop and roll out AI systems — both in their Northern homes and to the Rest of World — eliminating inefficiencies is being sold as a path to economic growth and success. Modernisation! If you want to be part of the global avant-garde, increase economic growth in your country and be part of the cool club. Rest of World policymakers are spoon fed a regular diet of this narrative that inefficiency is the nasty inhibitor of success. Inefficiency is the enemy of progress. Efficiency now!
But is it really?
Culture is not efficient. And I think we can agree that we all need culture.
Culture within any society creates the threads that bring people together. Culture is the fabric of the shared meaning that surrounds us: the symbols, stories, histories, values, rituals, foods, smells. All the things that help us make sense of the people, spaces and places around us. That makes our spaces unique, cherished.
Take for example the practice of gotong royong from the Indonesian and Malaysian experience. Taken from the Javanese language, gotong royong reflects a sense of mutual cooperation, bringing a community together, to help out, and achieve some common objective. It represents mutual cooperation — communities coming together to build, repair, and celebrate. It’s not necessarily fast nor cheap. But it is collaborative and often transformative, strengthening the solidarity that binds communities together.
I’ve participated in this firsthand in Aceh, Kintamani and other places across the Indonesian archipelago. Often the focus was a school, a community hall or village office – something the community decided upon. Then on appointed days, people show up to get dirty and build. Carry shingles and bags of cement. Cook meals for community members. Watch kids while parents volunteer — while community infrastructure rises from literally nothing.
From a project management lens, it’s a nightmare. From a human lens, it’s pretty much everything.
Anyone who has managed volunteers knows that this can be incredibly inefficient. People don’t show up on time, bring different skill sets and have various preferences and limitations. It's so much easier to outsource a contract to get things done. Quickly. Effectively. Efficiently.
But then you’d miss the point as to why we’d choose the inefficient approach over efficiency maximisation. Each seemingly inefficient input strengthens the social infrastructure that holds the community together. That’s why.
It seems counter-intuitive, right? That inefficiencies could lead to the strongest outcomes.
But that’s what gotong royong does. And this isn't unique to Indonesia. The Philippines has bayanihan, Korea has saemaul undong, and Thais embrace sanuk. Each celebrates collective effort, joy, and participation — all possibly quite inefficient. All utterly essential.
I asked my Filipino friend with whom I enjoy biking around town, and he described bayanihan as “Coming together = unity. It's a way of life that reflects our values of community, solidarity, and selflessness.” It's a tradition of helping others without expecting something in return. A Filipina friend described it as “the sense of community to accomplish something for each other. Knowing that they can rely on each other for aid.”
Did you catch that? The sense of accomplishing something FOR each other.
Similarly, once a community completes a community endeavour together, the rituals come. In Balinese Hindu culture, for example, the melaspas ceremony blesses a new space — purifies it, anchors it spiritually, and makes it ready for human use. Offerings, holy water, prayer — all “inefficient” by Silicon Valley standards. But essential by human ones.
What good is a community center without a sense of community? What good is a house without a sense of home? And how do you even turn those ideas into 1s and 0s for a computer? Should we even?
Let’s talk about sanuk for a bit. Thais generally share a worldview that emphasises enjoyment in every aspect of life. Whether work or hobby or pleasure or community building, activities should not only be productive, but also enjoyable, finding a balance between obligations and pleasure. And let's face it, if you’ve ever seen people being joyful or enjoying life – and hopefully you have, too – you know there is absolutely nothing efficient about it. Fun is not always efficient. No More Efficiency!
All totally inefficient. And, yet, all absolutely necessary to building strong social infrastructure.
So when big tech promises that AI will eliminate inefficiencies — in government, in business, in daily life — I get nervous.
Because what we call inefficiency often is the connective tissue of a healthy society. That’s where we build trust, maintain rituals, and express our identities.
AI systems that aim to optimise everything may sadly nudge us out of the behaviours that are core to our humanity in our diverse human contexts — volunteering, waiting, storytelling, pausing to laugh, delivering Yakult. We risk weakening the social infrastructure that the inefficiencies themselves are so incredibly important for constructing.
What happens when we no longer need to ask for directions? Ditch volunteers for more efficient gig workers? Stop attending ceremonies in person to watch shorts later on YouTube, or give up shared moments with strangers at the post office? Serendipity begins to disappear.
So, uh, what does this mean?
It means we can’t uncritically accept the narratives being sold to us — by tech companies, policy influencers, and the hype machines driving AI adoption. The idea that “efficiency” should override all other considerations is not neutral. It’s a worldview (of effective accelerationism, for example). And it’s one that leaves very little room for the beautifully messy, emotional, unpredictable parts of being human across so many diverse places in our world.
A world where our human emotions and culture are treated as inefficiencies – either through the AI engineers’ benign neglect or intentional technocolonialism – should unsettle us.
I think we are already feeling it. We complain about how life is becoming “impersonal.” We are sensing the subtle erosion of the human touches that once defined daily life which are now being streamlined away by digital efficiencies.
You’ve probably experienced it, too:
No need to ask a stranger for directions — Google Maps has it covered.
No need to talk to a bank teller — an AI chatbot will guide you (or frustrate you).
No need to hand over cash or even speak to a sales clerk — the smart checkout will handle it.
No need to volunteer — just donate online and feel good about it.
No need to attend a community meeting — just react to a post.
No need to write a condolence note — a broken heart emoji on Facebook or IG will do.
Every one of these “efficiencies” saves time. But something else gets lost in the transaction: the friction where trust is built, where stories are exchanged, where relationships form. Each time machines have nudged us to use one of these time-saving tools, don’t we risk losing something?
What to do
When I worked at the United Nations, we had a running joke: whenever someone proposed forming a “loose group” to address a tough issue, it usually meant nobody really wanted to deal with it. It was a way of appearing responsive without committing to real action.
We can’t afford that approach when it comes to AI.