It is a truth universally acknowledged that any sudden change to the design of a digital service or media product can be a bruising experience for all involved. I still bear the scars from my time as online editor of this newspaper in 2013, when some users of irishtimes.com reacted furiously to a redesign of its look and layout. Some of the issues they objected to were technical glitches that could be resolved quickly. Others were strategic alterations that were necessary for the future direction of the website. But in many cases the problem was simply that people don’t like change.
The big technology and social media companies have been navigating this minefield for years. Every significant new iteration of Facebook, Instagram or Apple’s iOS (and there have been many) has been greeted with a cacophony of boos and threats of mass boycotts. Inevitably, things settle down after a while, users adapt and the outrage machine moves on. Sometimes, however, the backlash has led to a quiet rollback of features that proved simply too unpopular to survive.
The latest object of digital dissatisfaction is OpenAI’s ChatGPT. From a standing start in late 2022, it became the fastest-growing consumer technology service in history, racking up hundreds of millions of regular users and integrating itself into workflows, schools and dinner-table conversations across the world. But with the release of GPT-5 on August 7th, the familiar pattern of furious reaction has played out once more.
In a column in The New Yorker last week headlined “What if AI doesn’t get much better than this?”, computer scientist and writer Cal Newport surveyed the fallout. He spoke to analysts who are sceptical of the claims made by the many AI boosters about how transformative the technology will really be. In particular, he examined the confident projections that steep growth in computing power will inevitably, within just a few years, lead to artificial general intelligence – a system capable of outperforming humans at most tasks.
Where does Mairead McGuinness’s shock withdrawal leave the presidential race?
Crash, part two: austerity bites and Brian Cowen’s Morning Ireland humiliation
Can ‘technocratic daddy’ Mark Carney solve Canada’s deep-rooted problems?
A handful of billionaires and a million artists in penury: big tech’s effect on culture, and what you can do about it
But the latest update offers little evidence of this evolutionary leap. The improvements between GPT-4 and GPT-5 are incremental rather than earth-shattering. Is it possible that the whole thing is being oversold?
Commentator and digital rights activist Cory Doctorow has been making similar arguments. He is particularly sceptical about the latest marketing buzzphrase: so-called agentic AI. This is the proposition that a suite of services will soon be able to carry out many tasks – booking flights, ordering groceries, planning holidays, arranging insurance renewals. To do that, however, would require the active co-operation of the providers of those services. Why, Doctorow asks, would an airline or a supermarket make it easy for AI-powered crawlers to carry out automated transactions on behalf of consumers? Their business models depend on nudging you toward specific outcomes – higher prices, add-ons, loyalty schemes. The last thing they want is a robotic middleman.
Doctorow may or may not be right. We are, after all, beginning to see deals between AI companies and established booking services such as Expedia or OpenTable. But it would be helpful if the travails of GPT-5 led to a temporary moratorium on both the dystopian and utopian visions of AI’s future. Instead, perhaps we should pay more attention to what’s actually happening right now. That is remarkable enough.
Data from analytics firms suggest that global search traffic has already begun to decline significantly, as people turn to conversational AI tools for quick answers. Many readers will recognise the personal experience of drifting away from Google and toward ChatGPT or its competitors. Google is scrambling to fight back with its own AI product. But if search is replaced, the implications are enormous. The entire business model of the internet – the mix of advertising and subscriptions that has underpinned the digital economy for a quarter of a century – could be upended.
Meanwhile, millions of people are incorporating AI tools into their everyday work. Emails, reports, slide decks, schedules, grant applications – the kind of administrative drudgery that once ate up hours of the working week is increasingly being outsourced, at least in part, to the machine. In offices and classrooms alike, AI has slipped into daily routines with a quiet inevitability.
One reason for the negative reaction to GPT-5 is that many of those users had grown comfortable with the quirks and limitations of GPT-4. They had developed strategies for getting the best out of it and were irritated when those routines were disrupted by the upgrade. OpenAI moved swiftly to address their concerns by reinstating the older versions for paying subscribers.
Other users complained the new model felt colder, more rational and less empathetic. They spoke of a kind of bereavement, fuelling fears that people, some of them more psychologically vulnerable than others, are forming unhealthy bonds with software that mimics human interaction.
We find ourselves in a curious place. On the one hand, the most grandiose predictions of imminent machine overlords or silicon utopias seem, at best, premature. On the other, the technology is already reshaping fundamental aspects of how we work, learn and communicate. The reinvention of the digital economy, the steady seepage of AI tools into everyday life, and the unsettling psychological implications of humans bonding with chatbots – all of this is happening now, not in some distant speculative future.
Perhaps the lesson, then, is the same one I learned back in 2013. People hate change, even when it is inevitable. But change does not necessarily arrive with a single flick of a switch. It can seep in slowly, reshaping habits and expectations almost before we notice. The real impact of AI may not be the sudden arrival of a godlike intelligence, but the gradual reconfiguration of how we go about the ordinary business of living.
And that is disorienting enough.
