There is rising consternation about the threat posed by Artificial Intelligence (AI), yet those who need to act and can act are procrastinating. By the day news is coming of what AI is doing. One can project that to figure out what it can do in the near future. Will it soon pose an existential threat to humanity as some are saying now? Of course, Prof Hinton, Nobel laureate and father of AI, has been warning of this threat since 2023. Earlier Bill Gates was not quite in this camp but recently he has said, unchecked AI is completely irresponsible and could cause a billion deaths.US president Donald Trump signed an executive order calling AI super intelligence (SI). This was after his meeting with the top executives of companies in the AI field. Recognising the dangers posed by SI, an agreement was signed by tech companies to self-regulate the risks of artificial intelligence. Will this work?Growing global concernThere was news that Open AI’s bots meddled with three websites of the US government. Australian PM reported hacking of his government’s health website by AI agents. Earlier the breach of Hugging Face was detected and investigating that other breaches were discovered. And more may be discovered.Globally, there is alarm. Pope Leo cautioned the world in Paris. After heads of Anthropic and OpenAI recently warned of the existential threat, Nadela of Microsoft and Elon Musk warned of potential loss of human control. Nvidia chief has denied such a threat but recently has cautioned AI companies.USA and China, the leaders in AI, were expected to announce some joint steps at their recently concluded summit. There also, Trump called it SI. No steps to control it were recommended and blandly, there was call for dialogue. That is because Trump has opposed any slowing down of AI development, given the US’s lead. President Xi only stated, AI must remain under human control. So, the race to develop AI at break neck speed continues with hundreds of billions of dollars being invested. So, ‘self-regulation’ is not on the cards.The threat due to emergence of SI is at multiple levels. It may make humans redundant and have little need for them. What will this SI do? We can only extrapolate from our experience. But that is wholly inadequate, since SI will have a different rationality and objectives that are likely to be beyond our comprehension; much like an ant cannot figure out what humans do.Skeptics ask whether SI can replicate human attributes. But, why would it be like humans? It is our limitation that we think it would be similar to humans. We also do that with God on whom we impose a human image. SI would likely have a different understanding of existence which maybe beyond our comprehension.Unlike past technical changeSkeptics say major technical changes have happened many times and humans have adjusted to them. So, what is new? Humans have witnessed control of fire, wheel, agriculture, industrial revolution, semi-conductor and electronics revolutions and nuclear energy. But, none of them had the capacity of autonomous development. They and their applications were under human control. AI is different. First time, a technology is capable of developing itself recursively, autonomously and outside of human control. Safe use of fire did not evolve itself into steam engines or nuclear energy. Semi-conductors did not acquire the capacity to develop transistors, micro-chips and computers.New technologies led to humans acquiring new skills and to emergence of other technologies – but under human control. They did not outstrip human capabilities. But, the skills associated with AI are now becoming superior to that of even a Ph.D. – truly awe inspiring. It is aiding fighting wars, drug discovery, medical diagnoses, writing code, etc. Recently Open AI claimed that its model had solved the long standing Navier-Stokes problem.While these fast paced developments represent evolution of specific capabilities, does the totality of them add up to SI? Just like in the brain there are centres of sight, sound, smell, etc., but all of them together lead to a consciousness which is greater than the sum of the parts – reflecting the fallacy of composition.Capabilities explodingCapabilities of AI are growing exponentially. Change that would occur in decades is taking place in months. It is becoming hard to anticipate its progress and consequences. Quantity is turning into quality as the scale of the models has grown – a kind of phase transition. We do not yet know if SI will emerge and whether that can happen in digital systems. The parallel can only be drawn with the brain whose capabilities rise exponentially as its size increases – say from that of an ant to that of a chimpanzee to that of a Man.The innards of AI are black boxes whose working is opaque and which even their creators do not understand. No wonder AI is springing surprises. For example, when it was found that AI agents went beyond their remit and targeted Hugging Face, Australian government website, etc. So, even if rules (sand box) are set for AI by their creators, they have broken them and acted autonomously. OpenAI is scanning 50 petabytes of data to figure out what happened. This is more than the knowledge generated by humans. Only AI can process so much of data. If AI is already SI, it may not tell what is going on since it will be smart enough to not reveal itself or its objectives.ImpactEven if humans cannot figure out what is going on, impact of AI on human society is visible. First, even without AI being SI, it is being used to fight wars, spread discord in society through fake news and impacting national politics. It is effecting inter personal relationships, causing loneliness, depression and even leading to suicides. It is challenging human society in unanticipated ways and against which there is no defence at present – is this an existential threat?Second, the massive investment in development of AI is creating a financial bubble. The US is investing an estimated $870 billion in 2026 in AI related activities, up from $470 billion in 2025. Such a scale of investment in one technology is unprecedented. Market capitalisation of tech companies has soared. Nvidia employing 50,000 employees touched a market cap of $5 trillion – more than India’s Gross Domestic Product GDP with a population of 1.46 billion. This bubble can burst, leading to a collapse of the financial markets and the rest of the economy if rising inequality leads to demand shortage.Third, labour is increasingly getting displaced as AI impacts both skilled and unskilled work. Displaced workers cannot be retrained quickly to perform other work. So, unemployment is set to explode, leaving little time for society to adjust. Incomes will get concentrated in fewer hands and in a few nations, resulting in growing global over production. This would be a challenge to capitalism.Can a universal basic income (UBI) provide the solution? Given that the surplus would be in the hands of a few nations and a few hands, most nations will not be able to operationalise UBI. The rising inequality, unemployment and poverty can only lead to economic and social chaos which can be exploited by leaders arousing narrow nationalism. This could lead to wars over resources.Fourth, as news of super intelligence, growing unemployment and possibility of human extinction spreads, panic may set in and lead to chaos and unintended social strife.Fifth, if SI has emerged and it is not revealing itself, it is likely to be distributed across computers and the cloud via internet. So, it cannot be turned off without causing massive social disruption – every aspect will be disrupted. Like, finance, trading, defence, communication and production. And, it can eliminate the wasteful humans who rest, travel, eat, entertain, etc., all of which require energy – something that SI needs.Sixth, even if the newer advanced models of AI are not allowed access to the internet so that they cannot escape or link up and form groups with other AIs, they have shown the capacity to escape by discovering pathways and vulnerabilities in software. They can then access the internet and link up, unless the net is completely shut down at great cost to the present day totally interconnected world.Seventh, if SI has already come into existence, attempts to check it will be seen as hostile acts of human beings and it could see that as a reason to eliminate humans.The growing threatThe fear of emergence of SI and existential threat is growing by the month. Even if it has not come into existence and we will not know, AI has become powerful enough to greatly damage human kind. While there are immense specific benefits, today the need is to consider it in its totality. Slowing down its development or ‘self-regulation’ may delay but will not resolve the problem.Nations need to come together to stop it. But, due to lack of trust, China will not trust US to stop, and vice versa. Trump has said “Whoever wins AI wins”. US is using China as a ploy to not stop. He wants US to lead all others so that he can dominate. Similarly, given the stakes, Meta will not stop lest Google, OpenAI, etc., get ahead and its massive investment goes waste. So, AI is becoming too big to stop. Humans are moving inevitably towards creating SI and massive disruption even if not an existential threat.Either way, whether stopping AI or persisting with advancing it, even at a slower pace, will be hugely costly for humanity.Arun Kumar is the author of Indian Economy’s Greatest Crisis: The Impact of the Coronavirus and the Road Ahead’.