What would Dr. Strangelove do with artificial intelligence?

By Max Hastings

I travelled a distance last week to attend an off-the-record lecture on artificial intelligence, delivered by the head of one of the West’s foremost intelligence agencies. Like many people, even those familiar with the defense and security world, I know next to nothing about AI and the revolution that will impact every life.

The trip proved not to be worth the gas. Our think tank audience was treated to a half hour of bromides about how important AI is going to be; the need to regulate and use it wisely. However, our speaker didn’t trouble to explain what AI will do, beyond the obvious of thinking autonomously. I wondered, as I left, whether they really know what it’s all about.

I start with this vignette so that if you feel as ignorant as me, you should be reassured, because you have plenty of company; yet all of us can grasp the fact that AI will play a key role in decision-making of every kind, commercial, governmental and military. On battlefields, almost all future weapons systems will have AI capability, for instance to synchronize drone swarms and to respond to incoming missiles with literally superhuman alacrity. There are fears that when AI urges optimal courses on commanders, it will promote escalation.

Some sages go much further in their predictions. Henry Kissinger, writing from the grave in a new book completed and co-authored by tech elder statesmen Eric Schmidt and Craig Mundie, argues that “immune from fear and favor, AI introduces a new possibility of objectivity in strategic decision-making.” Kissinger and his colleagues see the possibility that machines may behave more rationally, and thus better, than such human aggressors as Russian President Vladimir Putin.

Most of us are more wary, and indeed nervous. If we mistrust China’s President Xi Jinping and America’s President-elect Donald Trump to make wise decisions about peace or war, we are even more appalled by the notion of machines possessing power to usurp their judgments. Steven Feldstein of the Carnegie Endowment for International Peace says that AI “is helping to precipitate a new age of geopolitics,” which “will inevitably shape the functioning of democracy and governance.”

The growing power of chatbots will enable malign engineers vastly to influence political opinion-forming by broadcasting a torrent of propaganda as fact, further poisoning public discourse in the fashion already achieved by many right-wing outlets through social media.

Given that every day we see airline computer systems fail and government databanks hacked, it seems terrifying to contemplate a future in which a systems glitch might launch a nuclear missile. We simply do not trust the “experts,” who seek to reassure us that AI won’t be granted power to do this.

Many of us have seen Stanley Kubrick’s superlative 1964 satirical movie Dr. Strangelove, in which a mad American general launches a strategic bomber wing against the Soviet Union, only for the US president to discover that the Russians have created a doomsday machine, which will unleash massed nuclear weapons automatically, in the event of an American attack. The rest is the future, as they say.

Six decades on, we are obliged to rely, as humankind has in the past, upon the inhabitants of the planet retaining sufficient sense of self-preservation that they don’t build systems with autonomous power to destroy us all. Yet it seems reasonable to fear that Xi, Putin, Trump and Kim Jong Un themselves understand the finer points of the technology no better than I and perhaps you do.

I never forget the story — a matter of record, not legend — of Winston Churchill, approached by his chief scientific adviser in August 1941 and asked for authority to pursue an investigation of nuclear fission, with the ultimate ambition to create a bomb. This was the seed later brought to its terrible harvest by US wealth and power, through the Manhattan Project.

Churchill responded that he saw no need for a more powerful weapon, when the means of destruction already available seemed perfectly adequate: “Personally I am quite content with the existing explosives.” Since the scientists were so keen, however, he raised no objection to authorizing their researches. My point here is that, as a Victorian born in 1874, Churchill couldn’t comprehend the epochal significance of nuclear fission, and really only did so after the US bombings of Hiroshima and Nagasaki in 1945.

My own father wrote me a letter days after my birth in December 1945, which he presented to me when I was 21, anticipating that my generation would spend our lives haunted by the specter of The Bomb. In reality, it is extraordinary how cheerfully we have gotten through the last 79 years, frightened by many things but only briefly — during the 1962 Cuban Missile Crisis — really scared by the prospect of intercontinental nuclear war.

Today, we know this much: AI will prove as critical a game-changer in the affairs of the world as the creation of the atomic bomb. The authors of the Kissinger book, Genesis: Artificial Intelligence, Hope and the Human Spirit, between spasms of optimism, admit: “If AI emerges as a practically independent political, diplomatic, and military set of entities, that would force the exchange of the age-old balance of power for a new, uncharted disequilibrium. . . Such an . . .order could witness an internal implosion of societies and an uncontrollable explosion of external conflicts.”

The prophets of AI anticipate war ceasing to be a contest between rival armies, navies or air forces manned by humans, and becoming instead a struggle between rival machine systems. Courage, or lack of it, will cease to be a factor. The skills that matter won’t be those of a Top Gun pilot but instead those of software engineers who, long before hostilities commence, program the technology that does the fighting.

The Ukraine war has shown how far missiles and UAVs — drones — now dominate battlefields. If I was running weapons procurement for any nation, I would decline funding for any new combat aircraft, which wouldn’t enter service much before 2040, that required a crew.

I recall my sense of amazement 20 years ago in a hangar outside Kabul when, for the first time, I watched a former B-52 pilot fly a drone. I sensed then, and know now, that I was glimpsing the future.

At least the Manhattan Project was run by real people, even if some were as flawed as Robert Oppenheimer. It seems right for us to be scared by the prospect of machines directing our future armed conflicts, not to mention driving the operations of market makers and hedge fund managers. Feldstein observes, “the prospects for the accidental use of force and potential escalation are acute.” But AI is coming. Our enemies, are working furiously to advantage themselves. We have no choice save to keep up, which is what the Manhattan Project scientists told each other through three years of herculean labour to build The Bomb.

We must seek comfort in the fact that we have survived the first eight decades of the nuclear era, which many of our parents believed would be the end of us. Maybe our children will muddle through the AI era, which, like The Bomb, can never hereafter be wished away.