The word ‘arrival’ has a strange power in the technological imagination. It implies a threshold crossed, a frontier left behind and a final station. There, the train of progress stops and something wholly new begins. For decades, artificial intelligence has been spoken of in such terms. Many describe the moment when machines cease to be clever tools and become, in some profound way, our equals. Few have voiced this expectation more forcefully than Silicon Valley’s most visible prophets. Sam Altman of OpenAI has suggested we may see artificial general intelligence within the decade.9 According to Dario Amodei of Anthropic, the slope of progress points steeply towards systems that reason and plan at or beyond the human level.
For these men, and the companies they lead, AGI is imminent. The ‘myth of arrival’ thrives on such proclamations. It creates urgency and sales-worthy headlines. Most importantly, it attracts investment by making the future feel inevitable. Inevitability is intoxicating.
Science has heard declarations of arrival before. In 1989, chemists announced ‘cold fusion’, promising limitless energy from a laboratory bench, only for experiments around the world to fail to reproduce the result. Earlier generations believed space was filled with a mysterious ether that carried light, a theory overturned when experiments detected no such medium. Astronomers once mapped canals across Mars, convinced they were seeing the handiwork of an alien civilisation. Each claim arrived with confidence and headlines. Each later dissolved under closer scrutiny. The history of discovery suggests a simple rule: breakthroughs rarely arrive on schedule, and certainty tends to appear long before understanding.
For every prediction of AGI’s nearness, there is sober scepticism. Researchers such as Melanie Mitchell and Gary Marcus remind us that, however fluent, current models are still sophisticated engines of pattern prediction, not minds. Ilya Sutskever recently stated that ‘superalignment’—ensuring a superintelligent system does what we want—remains unsolved.
Others at DeepMind and MIT warn that claims of ‘arrival’ serve corporate interest as much as scientific truth. No consensus exists on when, or even whether, AGI will arrive. Timeline surveys swing wildly. Some say there is a 50 per cent chance of significant change within decades. Others conclude there is less than a 1 per cent likelihood of it occurring before mid-century. The fog is thick, and the myth thrives on lack of clarity.
Scholars and practitioners alike acknowledge that AGI has never had a precise, universally accepted technical definition. Broadly, it refers to an AI system capable of human-level reasoning and problem-solving across many domains. But unlike narrow systems such as speech recognition or image classification, general intelligence has no agreed benchmark. Proposed frameworks based on breadth of cognitive abilities, human-like performance profiles or collections of tasks remain contested, experimental or incomplete. In practice, many claims about AGI still rest on improvements in model size, computing power or benchmark performance, measures that often track engineering scale or commercial progress more than genuine human-level intelligence.
This elasticity matters because it changes what the term does. In investor pitches, policy debates and corporate strategy documents, AGI often functions less as a settled scientific target than as a narrative anchor: a way to attract capital, frame geopolitical competition and justify regulatory urgency. In that sense, AGI has become as much a political and economic category as a technical one, its meaning shaped by incentives and agendas as much as by empirical evidence.
As the horizon of AGI recedes and its definition stretches to fit each new headline, the more immediate transformation is taking place not in imagined minds yet to arrive, but in the changing value of intelligence itself: how it is priced, exchanged and woven into systems of power and profit.
Underlying the myth is more than just curiosity about machine minds. Intelligence, once seen as a deeply human trait, is now being turned into a product. You can rent a bit of eloquence with an API call or get a reasoning companion with a subscription. ‘Intelligence-as-a-service’ is quickly becoming a key business model of our time. This shift changes what we mean by thinking; it can now be measured in tokens, priced in dollars and tracked by the second. The value of thought, once linked to human experience, now depends on server costs and licencing fees. What does it mean, then, to be human when intelligence can be bought and sold like electricity? We might be expecting to see the ‘arrival’ as a single moment, instead of noticing that a shift is already occurring—a shift in how we define intelligence.
Arrival, then, is less a scientific prediction than a theatrical script being followed. It would be useful to remember that Silicon Valley speaks in prophetic terms because it is grandiose and persuasive, and because it attracts investors.
These tools are already changing how students write or doctors research. And over time, these changes to how we think may matter more than any sudden leap to superintelligence. The question, then, is not: ‘When will AGI arrive?’
It is: ‘What will become of us if intelligence itself becomes infrastructure, leased by the hour, optimised for engagement, detached from experience?’
This excerpt from ‘The New Divide’ by Jibu Elias has been published with permission from Westlands Books.