Hailed as both the saviour of workforces around the world as well as the digital spawn of Satan, AI’s entry into our everyday lives has been as polarising as Trump’s second term.
OpenAI launched ChatGPT in November 2022. Initially released as a “research preview”, this version (GPT-3.5) was what propelled AI onto the global radar. After just three years it would be naïve to think we’ve tamed the beast. A new AI lexicon has emerged, but that has only shown us how complex and nuanced the beast really is.
Last year we moved from generative AI to agentic AI, which is slowly but surely making search engine optimisation (SEO) – the bedrock of internet business – obsolete. If your company hasn’t done so already, switching to generative engine optimisation (GEO) and learning a new the algorithmic language is required.
We now know that AI dependency is a thing, especially for Gen Zs, many of whom are outsourcing their critical thinking to algorithms. We’ve also been introduced to AI psychosis, which has tragically led to suicide for some, who became emotionally dependant on their chatbots. Now we’re wading through AI slop, which proves the dead internet theory. (The theory posits that the internet has been dominated by AI, bots, and algorithmically curated content since around 2016, rather than authentic human activity.)
The internet of dark things
The Dark Forest, a science fiction novel by author Liu Cixin, has become a metaphor for the internet as a toxic arena where “digital predators” (bots, trolls, or surveillance technologies) lurk. The dead internet theory has paved the way.
In mid-2025 we crossed a significant line. Bot-generated content, and engagement, officially surpassed human internet creation and usage. Mahindra estimates that only 38.5% of internet traffic is human, while 61.5% is non-human. Synthetic content (AI-generated articles) has also eclipsed human-written articles. Roaming inside this synthetic swamp are automated bots: 37% are deemed malicious, while only 14% are “good bots”.
Thanks to AI, you now have an increasing number of synthetic influencers on social media, which most people can’t tell aren’t human.
AI boosts productivity – a half-truth
From the start, the promise of AI was utopian. It would free us from monotonous grunt work and allow us to pursue more creative endeavours our time-starved lives do not allow.
That narrative quickly changed to concern when AI did start replacing mundane and time-sapping entry-level jobs, but also jobs ripe for automation, across all areas of an organisation. Business owners embraced this cost saving, kick-starting AI redundancy, and in tandem, an HR existential crisis. (In February 2026, Block CEO Jack Dorsey announced the axing of 4 000 of the company’s 10 000 employees.)
“Learn how to work alongside AI” became the management mantra and many employees who survived the culling did so, unknowingly signing a Devil’s pact. “AI fatigue” was added to the lexicon.
Software design and coding were two tech industry jobs most at risk of being replaced by AI. One survivor, software engineer Siddhant Khare, wrote an essay titled AI Fatigue is Real and Nobody Talks About It, which laid bare this Devil’s pact. He wrote: “AI genuinely makes individual tasks faster. That's not a lie. What used to take me three hours now takes 45 minutes. But my days got harder. Not easier. Harder.
“The reason is simple once you see it, but it took me months to figure out. When each task takes less time, you don't do fewer tasks. You do more. Your capacity appears to expand, so the work expands to fill it. And then some. Your manager sees you shipping faster, so the expectations adjust. You see yourself shipping faster, so your own expectations adjust. The baseline moves.
“Before AI, I might spend a full day on one design problem. The pace was slow but the cognitive load was manageable. One problem. One day. Deep focus. Now? I might touch six different problems in a day. Each one ‘only takes an hour with AI’. But context-switching between six problems is brutally expensive for the human brain. The AI doesn't get tired between problems. I do.
“AI reduces the cost of production but increases the cost of coordination, review, and decision-making. And those costs fall entirely on the human. You became a reviewer and you didn't sign up for it.”
When agentic AI surfaced, workers were told that “task execution” would switch to “task stewardship”. Nobody considered what the human cost of that would be.
The cost and loss of spatial identity
While everyone is focused on the technological impact of AI on humanity, no one was considering the cultural/aesthetic impact of AI infrastructure, until Edwin Heathcote did. His article for the Financial Times titled Welcome to a Post-Human World: The Vast Anonymity of the Data Centre was equal parts thought-provoking and depressing.
He observes that, “Every age has its own distinctive building type. For the 19th century it was the railway station, for the 20th century it was the skyscraper. For the 21st century it looks like it’ll be the data centre”. He could be right. McKinsey estimates that tech companies plan to spend up to $7.9 trillion on data centres by 2030 – roughly double the value of the entire UK economy.
If the Victorians created aesthetically breathtaking train stations and factories, and 20th-century architects did the same with skyscrapers, then why are data centres so monolithically ugly?
Heathcote suggests that data centres are made for “machines and microchips, not humans”. They exist solely to house armies of blinking servers and mainframes, which in turn need to be supported by armies of back-up generators and cooling systems. This machine-to-machine love fest generates extreme temperatures and high noise levels: a physically challenging environment for humans. Those who oversee this forbidding environment are reminded daily that they are in a space built for machines, not humans.
The AI maths don’t math
In December 2025 IBM’s CEO Arvind Krishna shared his thoughts on the gap between the eyewatering capital expenditure and the as yet unproven return on investment of AI’s future.
Based on current costs, he calculated that if a single 1GW AI data centre costs $80 billion – and Big Tech plans to build 100 of them – that would cost $8 trillion in infrastructure, more than the entire semiconductor industry has earned in its lifetime. To finance that you’d need $800 billion in annual profit, just to cover interest. He observed that not even the $3 trillion giants make that. To complicate matters, AI data centres only last five years before hardware becomes obsolete, which means that the whole system needs to be rebuilt repeatedly.
Commenting on his calculations, one person said, “Right now, AI creates essays, images, summaries, and code. Useful? Sure. But $8 trillion useful? No one knows. The entire industry is betting on future productivity booms that haven’t materialised yet.
KEY TAKEAWAY
There’s no doubt that AI will prove to be not only disruptive, but transformative for humanity. That might sound overstated but already what started out as a handy workplace “co-pilot” has seeped into our socio-cultural fabric, and in turn into the political discourse, spreading disinformation and swaying opinion. In just three years the AI pendulum has swung from zealous praise to doomsday pessimism, and back again.
The pendulum will continue to swing back and forth and finally settle in the middle. But what does the middle ground between utopian and dystopian look like, and who controls whom? Humans or machines?
Dion Chang is the founder of Flux Trends and author of A Pawfect Life. For more trends as business strategy, visit www.fluxtrends.com
