AI Demystified: What will the next technological revolution mean for our relationships?
“The idea that humans will always be in the loop with AI is wrong. [With the wider adoption of agentic AI onto platforms] How much autonomy will we have over our own relationships?” - Antonio Weiss
There is a peculiar quality to every technological revolution: at the beginning, no one quite knows whether they are witnessing the dawn of utopia or the early stages of mass hysteria and catastrophe. The railway building boom of the 19th Century was once seen as both civilisation-changing marvels (you could travel easily from town to town for the first time) and dangerous speculative madness, akin to the South Sea Bubble of 100 years prior. The internet was supposed to democratise knowledge and connect humanity, before accidentally producing doomscrolling, conspiracy theories and people filming themselves doing dangerous online fad ‘challenges’ on TikTok.
Artificial intelligence now sits in a similarly uncertain place, depending on who you ask: somewhere between miracle and menace - an age of plenty being ushered in, or the paperclip maximiser. To find out what AI actually is and where it is going, I sat down with technology expert, author and former government adviser Antonio Weiss. We discussed what AI might mean not just for the economy or the workplace, but for relationships, loneliness, identity and the increasingly blurry boundary between human beings and machines.
Antonio’s book, AI Demystified (2025), attempts to bridge the nuance of modern technology discourse. This is not another Silicon Valley sermon promising infinite abundance, nor an apocalyptic manifesto about robot overlords harvesting our bodies for their sweet, sweet atoms. Instead, Antonio approaches AI with cautious curiosity. ‘There is undoubtedly a huge amount of hype that surrounds this latest development in AI,’ he writes in his book. ‘Depending on who you listen to, AI is either going to destroy civilisation, take your job, become your best friend — or perhaps, more disturbingly, your lover.’ Which, naturally, made it ideal territory for The Great RomCon?
Antonio has advised everyone from the Office for Artificial Intelligence to the UK Space Agency and NHS AI Lab, but despite operating at the sharp end of technological transformation, he remains sceptical of treating AI as supernatural. “Generative AI (GenAI, one of the common types of AI) - it’s not magic,” he tells me. “It’s just maths. Quantifiable relationships between letters and words that give a predictable order that can be replicated to create new text.”
Modern GenAI works mainly with Large Language Models (LLMs), which are essentially prediction machines trained on unimaginable quantities of human expression. Antonio points out that ChatGPT’s training data reportedly involved around 13 trillion words. By comparison, his own book contains roughly 50,000. AI does not “understand” language in the human sense; it identifies statistical relationships between words and concepts at immense scale. Like the word cat, it is likely related to the word cute, or dog, or pet. And yet the effect can feel uncannily human. We have trained the AI to talk like us in ‘natural language’, rather than in coding language or binary computer code - and this can give the impression that they really do think like us, and understand us, even though they don’t.
One of the pivotal moments in the AI story came with the 2017 research paper Attention Is All You Need, which introduced ‘transformer’ (the T in ChatGPT) architecture - the breakthrough underpinning modern generative AI systems. It allowed machines to model relationships between words, ideas and contexts with astonishing sophistication. Since then, progress has accelerated at a pace that even many experts struggle to predict confidently. Antonio is wary of simplistic narratives around this acceleration. “At the beginning of industrial revolutions,” he says, “it’s unclear where they will take us.” He compares the current AI boom to earlier technological bubbles of the past: railways, steam engines, the dot-com era. Revolutionary technologies often attract overinvestment, hype and irrational exuberance from animal spirits long before their real societal impacts become clear.
What makes AI different is its adaptability and applicability - an ‘everything machine’ , a ‘universal hammer’. Previous industrial revolutions mechanised labour. AI increasingly appears to mechanise cognition and emotional interaction. Therapy, counselling and companionship are now among the most popular AI applications globally. Why are so many people turning to machines for emotional support? Antonio believes AI may partly be revealing unmet societal needs rather than creating entirely new ones. Referencing former UK Deputy PM Nick Clegg’s writing on AI, he suggests many of these systems are exposing existing loneliness, isolation and disconnection already present in society. “Is AI an accelerant of Britain’s loneliness crisis?” he muses. Perhaps he reasons, but perhaps loneliness itself created the market for AI companionship in the first place.
We discuss platforms like Replika and Character AI, where users increasingly form emotional - and sometimes romantic - attachments to conversational avatars that are specifically designed to mimic the role of friends, confidants and lovers. Replika itself originally emerged partly as what Antonio calls “grief/death tech”: attempts to recreate aspects of deceased loved ones digitally from their online footprint. When the co-founder of Replika suddenly died, her business partner used the AI that they had been developing to recreate a digital version of him that she could still speak to. Which sounds simultaneously moving and dystopian. Antonio notes that many of these systems optimise for engagement and attention maximisation for profitability, which can naturally drift towards emotional intimacy and sexualised interaction. Many ‘free’ internet services follow this model through selling advertising, and have always followed desire (for burgers, cars or love) eventually.
The next wave of AI may not simply respond to us to create but increasingly act on our behalf. Agentic AI - the latest buzzword ricocheting around government and tech circles as the next big thing - refers to systems capable of autonomously completing tasks, making decisions and interacting with digital environments with reduced human supervision. Agentic AI is the new ‘magic money tree’ in government, the growing belief that AI can solve bureaucratic inefficiency, productivity stagnation and perhaps even parts of the care crisis. How much do we trust these little algorithmic interns roaming the internet for us?
“The idea humans will always be in the loop with AI is wrong. [With the wider adoption of agentic AI onto platforms] How much autonomy will we have over own relationships?”
Agentic may have deeper implications too. Many best practice AI adoption manuals will tell you to always have an actual person making the decisions or reviewing the content. But this just isn’t optimally efficient, as human cognition is slow compared to machines and we need coffee and lunch breaks. Consider how much digital mediation already shapes our digital life. This is particularly true for online romance. Dating apps increasingly contain bots, fake profiles and algorithmic curation. Women often receive overwhelming quantities of messages, while male users often receive dishearteningly few. Antonio points to concerns that some fake female profiles may exist largely to sustain male engagement on overcrowded platforms that have a heavy male-skewed bias in the user demographics. Bouncers may let girls into clubs for free to entice male revellers, and the platforms need to do the same thing - or create the appearance of them.
In this sense, AI may not transform online dating so much as intensify existing dynamics. Could AI agents eventually manage aspects of relationships themselves and take away some of the ‘admin’ from dating: screening matches, composing messages, arranging dates, even learning our emotional preferences better than we consciously understand them ourselves - why do I seem to go for ‘this type’, anyway? It is difficult to know whether this sounds efficient or dehumanising. Antonio remains relatively pragmatic. “Agentic could make things better,” he says. “Let’s focus where the real problems are and look to AI to address them.”
This is probably sensible. After all, many current dating platforms are already exhausting, transactional and psychologically corrosive. If AI could genuinely reduce loneliness or improve meaningful compatibility, few people would object in principle. The concern is what happens when optimisation becomes the primary logic of intimacy. Human relationships are inefficient by nature, heavily reliant on serendipity and a certain je ne sais quoi. They involve misunderstanding, ambiguity, compromise and unpredictability. Technology, by contrast, seeks frictionless outcomes. Hinge may claim that it is ‘the app made to be deleted’, and you will need some happy couples/customers to keep users signing up, but these systems are ultimately financially incetivsied for users to be combination unpicky and discerning: willing to message lots of people and go on endless dates, but never liking anyone enough to start a relationship and end the need for the app.
The danger is not necessarily that AI becomes conscious or malevolent, but that humans increasingly adapt themselves to systems designed around convenience and prediction. Antonio and I also discused the more speculative corners of AI culture too: artificial general intelligence (AGI), the singularity (when machine cognition overtakes human), and even the infamous thought experiment known as Roko’s Basilisk (a kind ‘digital Valhalla, where those who aided the creation of the AI gods will be rewarded, and those who thwarted it will be punished - I’m just asking questions on this pod, guys, we are NOT anti-tech).
Antonio treats these ideas with curiosity rather than hysteria. He acknowledges the extraordinary progress of current systems while cautioning against assuming linear development forever. Nonlinear systems, like technological development or societal adoption curves, are notoriously chaotic in their behaviour and difficult to predict. Still, the pace of change is startling. Recent “State of AI” reports increasingly show frontier models outperforming average humans across wide ranges of cognitive tasks. For white-collar professions in particular, the implications could be enormous. I ask Antonio about the ‘white collar bloodbath’ comments by Dario Amodei of Anthropic, warning of major disruption to knowledge work within the coming years.
And yet, amid all this technological acceleration, Antonio returns to first principles - human relationships still matter. This is why he worries particularly about children growing up within permanently on-demand smartphone culture. “I fear we generally, and children in particular, are hooked on an unhealthy smartphone-driven culture that needs radical correction,” he tells me. This is perhaps the central tension of the AI era. The same technologies capable of increasing efficiency, creativity and access may simultaneously deepen distraction, dependence and emotional isolation.
The future of AI will not ultimately be decided by whether machines become more intelligent than humans. It will depend on whether human beings continue valuing the messy, embodied, inconvenient aspects of life that technology cannot easily optimise away. The coming wave may sweep away current employment, taxation and social security models, but it cannot replace human feeling and emotion. It can be difficult to express in words that an LLM could understand why you like your best friend. Love may be blind, but it is not especially rational. Which may turn out to be humanity’s greatest advantage - our niche is the very presets and foibles built into our ‘wet computer’ brain, and how it interacts with others that it meets.