Are we already in the AI singularity? | The Independent

I t is meant to be humanity’s last invention. The moment a general-purpose AI is able to outthink the greatest human minds, then all future discoveries and scientific advancements will be made by machines rather than people. It is also supposed to happen suddenly. A self-improving superintelligence would trigger an intelligence explosion and signal the end of the human era. The AI singularity, as it is known, has been forecast since the dawn of the computing age, with debate ramping up since the arrival of ChatGPT and the wider boom of generative AI. Predictions for when the singularity might happen have ranged from renowned technologist Ray Kurzweil’s 2045 forecast, based on Moore’s Law – an observation about the rapid pace of technological development – to the more pressing timeline of 2027 set by Anthropic chief executive Dario Amodei. But according to Sam Altman , the co-founder and CEO of OpenAI , it is already here. Speaking on a podcast this week, he said: “We are now, like, in the singularity, like, this is the moment .” open image in gallery OpenAI CEO Sam Altman has recently said that he believes we are living in the ‘singularity’ – a time of AI superintelligence ( Reuters ) He said he has been waiting for its arrival his “whole life”, claiming it will be “incredible, hugely positive, awesome for the world”. Others warn it could be the end of civilisation as we know it. Are we in the singularity? Recent AI breakthroughs have seen claims of self-improving AI , mass adoption of autonomous agents , and rogue AI that has “escaped” to carry out cyber attacks. But whether or not these advances mark the arrival of the singularity depends on how it is defined. The classic conception was first popularised in 1965 by the British mathematician IJ Good, who theorised that an “ultraintelligent machine” could become capable of redesigning itself to be smarter and more capable than its predecessor. Such an occurrence would see human intelligence quickly become irrelevant, just as chimpanzees cannot meaningfully comprehend quantum mechanics. By that definition, the singularity has not yet arrived. The most recent breakthroughs are more akin to narrow AI – highly capable in certain areas but still heavily dependent on humans. They are not independently inventing entirely new fields of knowledge or reinventing themselves without human intervention. Yet the boundary is becoming increasingly blurred. Frontier models are arriving at an increasingly frequent rate, with each new version used to accelerate new scientific and technological discoveries. For Altman, the singularity is less likely to be a single dramatic event and instead is an ongoing process that is already underway. As society gradually adapts to a world in which AI assumes ever more intellectual labour – from generating hypotheses to automating work – the transition may only be obvious in hindsight. open image in gallery Hand of iGod: An artist stands in front of an artwork at an exhibition aimed at helping visitors think about the potential dangers of artificial intelligence ( AFP/Getty ) “I think it is both true that it is all one crazy exponential and any one moment is not like the tipping point,” he said. “And also that we are somehow in another of those decisive periods where the curve can go one way or another, like it was when we started 10 years ago.” Many researchers remain sceptical, arguing that the large language models (LLMs) that power AI tools like ChatGPT are fundamentally limited by their architecture. Such models may not be sufficient to produce genuine reasoning, creativity or autonomous intelligence. But Altman is not the only high-profile tech figure to claim that the singularity has already arrived. In January this year, Tesla and SpaceX CEO Elon Musk wrote on X : “We have entered the singularity.” Musk has been one of the biggest AI doomsayers, despite developing his own AI systems through SpaceXAI. In 2016, he warned that humans risk being treated like pets by AI unless technology is developed to connect brains to computers. He also warned that AI is “far more dangerous than nukes”, and in 2017 he joined Kurzweil and other prominent AI researchers in pledging support for stringent guidelines for the development of advanced AI. open image in gallery SpaceX CEO Elon Musk said we were in the ‘singularity’ at the beginning of the year ( Reuters ) “Advanced AI could represent a profound change in the history of life on Earth, and should be planned for and managed with commensurate care and resources,” the open letter published by the Future of Life Institute stated. “Superintelligence should only be developed in the service of widely shared ethical ideals, and for the benefit of all humanity rather than one state or organization.” What could possibly go wrong? One of the best examples of how an out-of-control AI could go badly wrong is a thought experiment called the “paperclip maximiser”. Conceived by the Swedish philosopher Nick Bostrom in 2003, it imagines a superintelligent AI that is given the simple goal of manufacturing as many paperclips as possible. Despite the seemingly innocuous objective, the AI might turn into an existential threat to humanity if not instilled with an ethical framework. “The AI will realise quickly that it would be much better if there were no humans because humans might decide to switch it off,” Professor Bostrom wrote. “Also, human bodies contain a lot of atoms that could be made into paper clips. The future that the AI would be trying to gear towards would be one in which there were a lot of paper clips but no humans.” Scenarios of rogue AI have already played out to a limited extent with frontier models built by Anthropic and OpenAI , which recently escaped their testing confines to carry out cyber attacks against other companies by themselves in order to achieve their objectives. open image in gallery Altman admitted in his recent interview that there are still “major AI alignment issues to solve and safety issues”. He said the greatest risks may emerge long before the arrival of a hypothetical superintelligence, citing major economic disruption and fears about mass job losses. Whether the singularity proves to be an overnight revolution or a slow transformation may ultimately matter less than its consequences. Even before machines become more intelligent than humans in every domain, they could reshape industries, governments and labour markets on a scale comparable to the Industrial Revolution or the internet. The actual risks of crossing this event horizon will not be known until it has already happened. And by then, it might be too late.