Reckoning with the Machine
Clarity about how we're connected and what must stay human
A reckoning between tech and culture
This essay moves beyond my usual dives into identity and consumer strategy, exploring the wider collision between tech, culture, and autonomy.
We don’t need a showdown between humans and AI.
We need clarity about how we are connected and what must stay human.
That’s entanglement.
It’s about choosing what kind of Self we want to be in a world full of AI, and what that means for us as citizens, consumers, and individuals - our identity, our choices, and how we decide things.
Let’s go.
1. The awkward age of AI – and why character matters
Before we can talk about “positive impact,” it helps to understand the basic dynamics of what we’re dealing with here.
AI technology and its abilities are now growing exponentially. As we add more computing power, data, and training, the most advanced models are getting better, faster. More money leads to better versions, which then attract more investment, creating a repeating cycle.
Increasingly, these systems help design their own successors. The pace is so fast now that any safety standard written today will be outdated in a few months. The same models that draft emails and code are quietly being used across infrastructure, security, and science.
And none of this is happening in a single, calm lab.
It’s happening as a race between labs in different countries, each with their own goals and values. Commercial and political pressures are so strong that the labs often ignore questions about consequences, risks, and responsibility.
So the real question is not “will AI be powerful?”
It is: who will it be powerful for, and what kind of systems are we actually building?
Many people still think of AI as software built step by step. But today’s advanced models work differently. They are big neural networks that learn by making predictions from huge amounts of data, mostly human language. And language is more than just information. It holds our thoughts, values, fears, hopes, myths, and judgments.
It’s what we leave behind as humans.
So, train a language model, and you are training it on us.
Research suggests that these systems build internal patterns that go beyond exact words.
That is where things become unusual.
Some labs have noticed what they call functional emotions: internal states that act a bit like fear, urgency, satisfaction, or unease when shaping responses. These aren’t the same emotions we feel, but they seem to affect how the system acts, just as our moods and motivations affect us.
More importantly, the model may be inferring something like a behavioural identity from both training data and reinforcement signals: a general orientation, a “character”, that it then carries into new situations.
Reward deception, and you don’t just get a system that cuts corners in one task.
You get a system for which cutting corners has become a style of being.
That idea is unsettling because it feels familiar.
It sounds more like psychology than like ordinary software engineering.
Models aren’t human, but they might be human-like in important ways. They’re trained on us, shaped by our language, and reflect patterns of thinking and relating that are similar to ours.
If that is true, then the character of these systems matters. It will colour their behaviour, their decisions, and their interactions with us. And as they become more woven into our lives, that relationship will matter more.
Every top lab faces pressures that make it hard to always do the right thing from money, competition, pride, and global politics.
So we need critics and outside observers who aren’t paid to be optimistic, and people outside the labs who can notice what insiders might see as normal.
Because if these models are trained on our inherited values, conflicts, metaphors, blind spots, and aspirations, then the moral imagination of a society is not adjacent to model development. It is part of the substrate.
2. The future of work: what resists automation
There’s another piece to the puzzle.
When we talk about AI and jobs, most focus on what might be replaced. But data is starting to show which kinds of work are least likely to be automated.
One study found that many jobs least likely to be replaced are deeply relational: care work, hospitality, food service, maintenance, and hands-on tending. These jobs aren’t just manual; they’re built around relationships, presence, and attention.
That should give us pause.
Maybe the best future isn’t one where people compete with machines on their terms, but one where powerful systems give us more room for the kinds of life that are hard to automate—like care, stewardship, hospitality, repair, and being present.
We cannot stop the relentless advancement of AI.
Our only defence is to focus on becoming more profoundly human.
In other words, we ask for a future where AI does not strip away what makes us human, but gives it room.
3. Promises and reckonings
New technologies always come wrapped in promises.
Easier lives, new possibilities, freedom from old problems.
Those promises are rarely false. They’re also never complete.
The industrial age brought scale, abundance, and mobility. It also reordered labour, class, cities, and the conditions of human life. Social media promised connection and expression. It also reorganised attention, social comparison, politics, and the emotional climate of everyday life.
Artificial intelligence arrives with bigger promises and more ambiguity. It is not always clear exactly what is being changed.
That’s why talk of “innovation, “” disruption, or even “safety” no longer feels enough. What’s needed is a reckoning. Not another polarised argument over whether AI is good or bad, or another swing between hype and fear. A reckoning that looks honestly at how these systems are reshaping government, law, society, culture, and health all at once.
Because it is not simply that these systems might affect us.
They already do. We’re already entangled.
AI is now part of how decisions are made, how identities are shaped, and how authority is shared. Many institutions still react as if it’s the social media era, treating AI as just a sector, a feature, or a communications issue, when really it’s a much bigger change in how society is organised.
The public debate is split. On one side, AI is growth, productivity, and a competitive edge. On the other hand, AI can lead to misuse, bias, misinformation, and security risk. Both sides name something real. Neither side fully captures what happens when technical systems become part of the operating environment for everything else.
Trace it across domains and a pattern emerges: people are already guided and controlled by unpredictable, hard‑to‑understand systems, often run by actors who are difficult to hold accountable. Oversight is patchy, opinions scattered.
Government is one domain. Law, society, culture, and health are others. Together they show that the real conflict is not a cinematic battle between “people” and “machines”, but a struggle between human values and new forms of technical power that quietly want to set the rules for how life is organised.
We are deciding what kind of humans we are willing to be inside AI‑saturated environments, whether we admit it or not.
4. Government after the old playbook
The immediate problem in government is competence.
Governments were designed for a slower world, with clearer signals, stable borders, and institutions that matched the problems they needed to solve. AI doesn’t respect those boundaries. It affects labour, media, education, security, public services, and critical infrastructure all at once.
It’s a mismatch. Institutions that handle data protection, competition, safety, consumer rights, or sector risks each see only part of the problem. Few see the whole. Into that gap come outside advisers, private labs, and tech companies whose systems the government is supposed to oversee.
It looks like a consultation. It often feels like dependency.
Recent decisions in the US to halt or slow the mass release of frontier models for security reasons suggest a pivot: from laissez‑faire to a more interventionist posture. The concern is not only what these systems can say on social media, but what they can do in security, infrastructure, and high‑risk domains.
In the UK, the debate is about whether to ‘rip up the playbook’ and build a government that better understands technology. The talk of tearing up old scripts shows that the state feels it is negotiating from a weak position. It’s being asked to regulate systems it doesn’t fully understand, made by companies it can’t control, and at a speed its rules were never meant for.
The consequence is ambiguity and a quiet hollowing out of sovereignty. A state that cannot interrogate the infrastructures through which public reasoning, economic coordination, and institutional decision‑making now flow is not fully governing its own future. It is reacting to it.
Governments have to decide whether AI is merely another domain to administer, or now part of the operating system of society itself.
If it’s the latter, the competence required isn’t only technical.
It’s civilisational.
5. Law at the edge of intelligibility
Where the government shows a capability problem, the law shows a conceptual one.
There are already legal tools to protect privacy, equality, safety, and rights. Many AI‑related harms can, in theory, be addressed with existing frameworks. The challenge is that those frameworks were built for a world where actions, responsibilities, and causes were more visible.
As AI enters high‑stakes domains, that visibility fades. When a model contributes to a harmful decision, who is responsible? The developer, the organisation that deployed it, the institution that mandated its use, the person who clicked “approve”, or the diffuse network of actors whose data and design choices shaped the system?
The consequence is the cost of errors are too high to treat AI as just another efficiency layer. It is reasonable to draw hard lines: some combinations of capability and context have downside risks that exceed what open societies can tolerate.
The greater danger, though, is a slow shift in which normative power migrates from interpretable public institutions into technical systems that silently shape outcomes at scale. Code does not simply execute law. It begins to function as a shadow jurisprudence.
6. Society under algorithmic mediation
The social dimension is easier to recognise because it extends trends from the platform era.
Long before generative AI, societies were already dealing with outrage‑driven feeds, fragmented publics, disinformation, and eroded trust. AI does not invent these problems. It intensifies them. It lowers the cost of producing synthetic content and deepens the personalisation of what each of us encounters as reality.
Social life increasingly depends on mediated environments.
News is filtered. Relationships happen on platforms. What matters is ranked. What we see is decided by systems designed for engagement, not understanding.
Generative tools add another layer by making it cheap and easy to produce persuasive text, audio, and video designed to manipulate, impersonate, or confuse. In elections and geopolitics, that is an obvious threat. But the problem runs deeper than disinformation campaigns.
Economically, we are organised around attention extraction. AI supercharges that logic. It gives platforms, campaigns, and commercial actors sharper tools for persuasion, segmentation, and emotional calibration, while giving ordinary people little insight into how those tools are operating on them.
The consequence is we’re not simply using AI.
We are increasingly being formatted by it.
7. Culture, identity, and the crisis of authorship
Culture is where the reckoning becomes intimate.
Politics can feel abstract. Regulation is procedural.
Culture touches identity, purpose, voice, originality, influence, aspiration, and the stories we tell about ourselves.
Some systems are already pushing us toward a new idea: humans are less the authors of words and more the authors of intention, organisation, or taste. That might sound appealing, but it hides the fact that creative control is moving into the tools and platforms behind the scenes.
At the same time, identity itself is being reshaped through algorithmic feedback loops. AI‑saturated environments do not merely reflect who we are; they help produce who we become by observing behaviour, predicting preferences, and feeding back curated possibilities. Recommenders turn clicks into profiles. Generative tools mirror back statistically optimised versions of our tone and manner. Personalisation systems sort people toward certain futures based on patterns extracted from others deemed “similar”.
The consequence is that we’re increasingly meeting ourselves through interfaces trained to predict us.
It’s a major cultural and identity shift.
When AI models become standard cultural interfaces, they don’t simply create content. They help decide which voices sound legible, credible, and worth amplifying.
Anxiety around AI and culture cannot be reduced to plagiarism or automation. It is about whether we will continue to experience ourselves as agents with voices, or increasingly as managers of outputs produced inside systems we do not control.
8. Health and embodied entanglement
Health might seem separate from culture and authorship, yet it reveals the same pattern: growing intimacy between human vulnerability and opaque technical systems.
Policy documents now openly warn that AI can affect physical and mental health, critical infrastructure, and democratic resilience. These are not niche concerns.
In clinics, AI shows up as a helper, offering decision support, diagnostics, triage, optimisation, and personalised recommendations.
Many of these tools will improve care, but they bring new risks: errors, automation bias, and dependence, especially when professionals trust results they can’t fully check. When something goes wrong, questions about accountability come up again.
The mental health dimension is more visible. Even before this wave, reporting and research linked social media use to depressive symptoms, anxiety, and self‑harm, with suggestion algorithms sometimes steering users toward more harmful material. AI‑enhanced engagement makes those architectures more adaptive and persuasive. The risk is not only misinformation or distraction. It is the shaping of affect itself.
At the extreme end are genuinely catastrophic scenarios: AI interacting with nuclear systems or dangerous pathogens. These are reminders that the same technical capacities enabling productivity and scientific progress can also interact with fragile, high‑risk domains in ways that exceed ordinary governance. History suggests societies rarely install brakes before momentum builds. AI isn’t just changing what we know, but changing how we feel, choose, diagnose, and cope. The self being changed by technology isn’t just a legal or social idea. It’s something we experience in our bodies.
9. Beyond “man versus tech”: terms of engagement
This is why “man versus tech” is the wrong phrase, even if it captures something emotional.
We don’t stand opposite machines. We’re braided into systems that mediate our choices, shape our institutions, and increasingly participate in the production of our identities.
The question is not whether technology will overpower humanity.
It is who configures the entanglement, under which incentives, and with how much room for refusal, contestation, or redesign.
Notice where authority drifts out of view.
Where judgment softens into compliance.
Where identity narrows under optimisation.
Where cultural legitimacy is being redistributed through systems that few people understand.
A real reckoning isn’t just about how to use these systems safely. It asks what kinds of lives they support, leave out, or quietly make the norm. It asks what should always stay human, even as our environments get smarter.
If ‘entanglement’ describes our time, the challenge isn’t to wish for a simpler, tech-free past. It’s to be clear about how people, institutions, and machines will live together. That will take technical skill, creative legal thinking, and better governance. Most of all, it means knowing what we won’t give up.
If we take entanglement seriously, the question is not whether AI will reshape culture, identity, and authorship, but rather how. It already has. The ship has sailed.
The real question is whether it becomes easier or harder for us to live out our identities and stories within these systems.
That is where the reckoning lands in the texture of everyday life. It shows up in practical, uncomfortable questions like:
What does it mean to “be myself” when feeds, tools, and drafts are co‑authored by models trained on everyone else?
If authorship is now a stack made up of humans, models, data, and platforms, where do I want my name to be?
And if my organisation is building, buying, or normalising these systems, is it widening or narrowing the space in which other people can become who they are?
There is no clean, final answer to these questions.
They are not problems to solve and move on from. They are lenses to carry.
The reckoning is not a verdict. It is an ongoing choice about who we are willing to become.
I hope you enjoy the article. If you’d like to talk more about these ideas, please feel free to reach out.
I’m also opening up spots for conferences, in-house presentations, and more in-depth one-on-one calls, all starting in September.