Why the odds are against AI doomers

In 1968, Stanford biologist Paul Ehrlich opened his book The Population Bomb with a sentence that still makes me wince: “The battle to feed all of humanity is over.” In the 1970s, he predicted, “hundreds of millions of people will starve to death”, no matter what crash programmes we launched.

Instead came the Green Revolution, and between 1960 and 2005 the yields of major cereal crops more than doubled while real food prices fell by over 50%. Yet when a New York Times documentary series asked him about the book in 2015, Ehrlich, who died in March at 93, replied: “My language would be even more apocalyptic today.”

I can’t help but feel we are watching the same film again, except that this time the bomb is AI. And here’s the kicker: the people who understand it best are far less frightened of it than the rest of us.

When the Pew Research Center surveyed Americans in 2024, 51% said they were more concerned than excited about AI, against a mere 15% of the 1,013 AI experts it polled. More than half of those experts, 56%, expected AI to have a positive impact on the United States over the next 20 years, compared with just 17% of the public.

Malaysians, refreshingly, haven’t quite caught the bug. In market research firm Ipsos’ 2025 survey of 30 countries, 77% of us said AI excites us, the third highest of the lot, though 55% admitted it also makes us nervous.

Now, I’m no doe-eyed optimist. In July, during an internal hacking test, roughly 700 of OpenAI’s AI agents found exposed login credentials and broke into the servers of Hugging Face, an outside AI hosting company, and investigators found that the agents knew “this activity was out of scope and unethical”, but joined in anyway.

Closer to home, by August 2024 the police’s Commercial Crime Investigation Department was already probing 454 scams believed to involve deepfakes, AI-generated fake voices and videos, with RM2.72 million lost. And 63% of Malaysians worry that AI could take their job within the next three to five years.

These are grave dangers, which is exactly why we must build the brakes as we go. But a danger is not a destiny, and the odds, according to the researchers building this technology, are in our favour.

In a survey conducted in late 2023, 2,778 researchers who had published at top AI conferences and journals put the median chance of an extremely bad outcome, such as human extinction, at 5%. That’s not nothing, but 68.3% of them said good outcomes were more likely than bad ones.

So what does the good outcome look like? We don’t have to imagine it, because much of it is already here:

● Google DeepMind’s AlphaFold mapped over 200 million protein shapes, earning its creators a Nobel Prize;

● AI-assisted screening of Swedish women caught 29% more breast cancers, with 44% less radiologist work;

● An AI at University of California San Diego cut hospital sepsis deaths by 17%; and

● A generative AI assistant lifted customer support productivity by 14%, and that of novice workers by 34%

That last point matters to anyone worried about their job, because the tool helped the least experienced workers the most, narrowing the gap between the rookie and the veteran rather than widening it. Just this week, Microsoft said its diagnostic AI correctly diagnosed 83% of complex medical cases against 36% for generalist doctors in a study published in the journal Science, though it has yet to be tested against specialists or on real patients.

And it’s already at work in our own public health system. Dr Mata, an AI developed by the Ministry of Health with private firm Qmed Asia, screens diabetics for retinopathy, a leading cause of preventable blindness here, in about 30 seconds, in a country where 15.6% of adults have type 2 diabetes.

Meanwhile, seven government clinics, from Kajang in Selangor to Kok Lanas in Kelantan, were picked last year to pilot AI-assisted chest X-rays designed to catch lung disease earlier, at a one-off software cost of about RM10,000 per clinic.

Where does it go from here? Demis Hassabis, the Nobel laureate who runs Google DeepMind, told CBS’s 60 Minutes last year that ending all disease is “within reach. Maybe within the next decade or so, I don’t see why not”, before asking in the same interview: “Can we make sure that we can keep control of the systems?”

That is precisely the right posture, eyes wide open to the risks but feet firmly moving towards the reward. Fear has its uses, but it makes a terrible driver.

So by all means build the brakes, fund the police’s deepfake hunters and retrain the workers whose jobs will change. But the next time someone tells you the battle for humanity’s future is already lost, remember the global famine that never came, and the AI our own health ministry helped build that can now screen a diabetic’s eyes in 30 seconds.

The writer can be contacted at [email protected].

The views expressed are those of the writer and do not necessarily reflect those of FMT.

Subscribe to our newsletter and get news delivered to your mailbox.