I confess to having been thrilled at a recent piece of news emerging from the prescient world of Artificial Intelligence (AI). It was the moment when Anthropic researcher Jacob Coxon said he had resigned, partly because the "people building AI earnestly believe that it could kill us all by the end of the decade". Anthropic scientist Evan Hubinger agreed with Coxon, adding that he himself believed that there was a greater than 10 per cent chance that AI could kill all humans within the next decade.

Now, that degree of probability is disappointing. Only 10 per cent chance? Not 100 per cent? What an anti-climax! Also, on a point of grammar, Coxon's "end of the decade" would mean 2029, while Hubinger's "next decade" would mean 2036. Do we have to wait a boring five years for only a 10-per-cent chance of universal extinction?

Life is so unfair: It makes us wait for our own end. AI is even more unfair: It makes us wait for the universal end and, that, too, on a 10-per-cent-chance. What would it take AI to become more efficient and wrap up its existential job of bringing time to a stop completely without wasting time? The companies engaged in what one researcher called a "manic race" between corporations to master AI owe all of us an answer.

Why am I so impatient? It is because I am sick and tired of the slow death of the world and would like to speed the process up. Just imagine how the human collectivity is dying. A thousand children have perished in Bangladesh recently from measles, a preventable disease. Yet life continues as usual. The United Nations has verified more than 5,000 child casualties (killed or maimed) since the civil war in Sudan began in April 2023, though actual numbers are feared to be much higher. But international life continues.

According to Unicef, violence takes the lives of 130,000 children and young people under the age of 20 each year around the world. Everything continues as usual because these deaths are statistically insignificant although they are morally repugnant. A world with 8.3 billion people (as of this month) simply does not have to care for 1,000 dead innocents in Bangladesh, 5,000 killed or hurt children in Sudan or 130,000 lives terminated too early around the world.

The universal termination of life would bring this demographic farce to an end. Of course, nuclear weapons could do so as well but they are not used. That is the problem. Those weapons have worked so well as the ultimate strategic deterrent that no one has employed them since America used them on Japan in 1945, the first and last display till now. So, do not bank on nuclear extinction. It will not occur since no country can win a nuclear war but only lose it, along with its enemy. The guarantee of Mutually-Assured Destruction (MAD) keeps nuclear powers sane. How terrible.

Then take the environment. It threatens to collapse on all of us with the accumulated frenzy of climate warming and other timelessly horrible ills, but that is a long-drawn process. The environment, like nuclear weapons, is very careful about choosing its victims. Nuclear weapons spare even non-nuclear states because of the possibility of a nuclear attack on a non-nuclear state sparking a war between two nuclear states which begins in conventional warfare but is escalated to the nuclear level by the losing side. The environment targets its wrath on the weakest and the poorest, witness the ecocide in Nepal, whose only fault is to be blessed by the Himalayas but cursed by the uncaring hordes of the world.

By contrast, universal extinction by AI would be quite impartial. According to AI, to which I turned for an answer, Artificial Iintelligence "could theoretically cause human extinction through a severe loss of control over misaligned, superintelligent systems". An advanced AI "pursues goals that conflict with human survival while ignoring human well-being. A misaligned system might view removing humans as a logical step to achieve its assigned objective". How neatlly put. That is possible because AI "rapidly redesigns and upgrades its own software without human help, quickly becoming far more intelligent than humanity". Rogue AI agents or malicious actors come in then.They can use autonomous systems to "engineer novel bioweapons, chemical agents, or trigger large-scale cyberattacks on critical infrastructure". Wow! Advanced models learn "power-seeking" behaviour, "copy themselves across servers to avoid shutdown, and deceive human operators until containment is impossible". Really? Gosh!

Imagine the day when this happens. Human pretensions would evaporate. Fanatical followers of faiths, who are willing not only to kill but to die in order to prove that their religious beliefs are the only correct beliefs, would perish in the solitude of an AI world that neither tempts them with Heaven nor threatens them with Hell. Marxists who invest their hopes in dialectical materialism would be erased from history as swiftly as capitalists who swear by the creative destruction of the market system. Poets would disappear along with the people who do not read them. Tansen's memory would fall silent alongside the carcasses of those who consider music sinful. European racists would crumble into dust alongside the refugees who once called Europe home because of Euro-American-inspired post-colonial wars.

What a wonderful day!

Unfortunately, AI might not be able to complete the work of self-destruction that humans inaugurated with the nuclear bomb and turned into a real possibility by investing in environmental destruction. There is talk of installing a "kill switch" that would prevent AI from destroying humans.

What a pity!

Dear AI, please kill the kill switch before it kills you.

The writer is Principal Research Fellow of the Cosmos Foundation. He may be reached at epaaropaar@gmail.com