Next Generation Medicine

The Real Singularity: AI, Medicine and the Quiet Win

AI will replace 80% of doctors. Radiologists? Time to find a new career.

That was the AI singularity doomsday message broadcast over the last decade, a message driven by some of tech's biggest hitters.

In 2016, Geoffrey Hinton, a founding figure of modern AI, proclaimed it was time to stop training radiologists, because deep learning would outread them within five years. Vinod Khosla, co-founder of Sun Microsystems and an early OpenAI investor, predicted in 2012 that technology would replace eighty percent of doctors.

The message to the white coats was simple: adapt, retrain or quit, because you WILL be automated.

The short version? None of it happened. Ten years on:

  • AI has not eliminated any medical specialty.
  • Radiology is growing, not shrinking.

So has AI in medicine failed? Not at all. The revolution has just not been televised. The wins have landed quietly in research:

This quiet singularity means AI has the potential to become the most powerful instrument in the doctor's bag. Those who learn to direct AI, choosing what to trust and what to override, will define the next decade.

Read on for more detail on reasons to be bullish on AI in medicine, and what it could mean for medical careers, clinicians and global healthspan improvements.

The prince that was promised

The promise came in two parts. AI would replace the doctors, and it would slash the costs. Reality has been blunt with both.

Take replacement first. Radiology was the specialty expected to suffer from AI, yet it has grown instead. Mayo Clinic's radiology staff has grown 55 percent since 2016. The American College of Radiology projects a further 26 percent by 2055, and demand for imaging keeps outpacing the supply of radiologists. Hinton himself, interviewed by the New York Times in 2025, conceded he had been wrong about the timing.

Costs fared no better. A 2023 McKinsey analysis with Harvard economist David Cutler confidently estimated potential AI-led savings at between $200 billion and $360 billion per year within five years. Three years into that window, these savings have not yet appeared. U.S health spending instead climbed to a record $4.9 trillion.

But writing the technology off completely at this point could repeat an old mistake. 

We have been here before

The mistake goes like this. A breakthrough misses the promised timeline, so we write it off just as its slower, broader impact gains momentum.

It has happened often enough to have a name. Amara's Law: that we overestimate a technology in the short run, and underestimate it in the long run. It is not limited to medicine.

The internet was declared the future in 1999, crashed in 2000, then spent twenty quiet years remaking the entire economy. The potential was real. The timeline was fiction.

Medicine has lived through it twice in a generation. The Human Genome Project was completed in 2003 amid confident talk of curing disease within the decade. It underdelivered for years, and is only now becoming routine clinical infrastructure through precision and longevity medicine.

Gene therapy was written off almost entirely after a patient died in a 1999 trial, only to return stronger two decades later with gene therapies for inherited blindness, spinal muscular atrophy and sickle cell disease.

In each case, the champions were too optimistic and the doubters were looking in the wrong place. Are we doing it again, this time with AI and the future of medicine? 

The quiet win

Researchers and scientists quietly harnessed AI to tackle some of biology's hardest problems.

AlphaFold has predicted the structures of more than 200 million proteins, effectively the entire known proteome. Its creators, Demis Hassabis and John Jumper, won the 2024 Nobel Prize in Chemistry.

Since a protein's shape determines how drugs bind to it, mapping nearly all of them handed drug discovery a new foundation.

Insilico Medicine used AI to find the target for rentosertib and generative AI to design the molecule. This is the first AI-assisted drug to reach patients, now being developed as a anr idiopathic pulmonary fibrosis treatment. In a Phase 2a trial, patients on the higher dose gained lung function while the placebo group declined.

An MIT team screened over 100 million compounds with AI, resulting in the identification of halicin, a structurally new antibiotic that kills superbugs defined by the World Health Organization as critical. They then used this foundation to design fresh molecules against drug-resistant gonorrhoea and MRSA.

The longer run

It now seems likely that over the next ten years, AI's key use will be in finding and designing drugs, not in treating patients. Two uses are already underway, and a third sits further out.

The first clear use case lies in developing the drug pipeline. Initial AI-discovered drugs have reached late-stage trials, and the early signals are promising, but Phase III is where drugs die. This stage will be the real test of any AI discovery advantage.

By the end of the decade, we should know whether AI-assisted drug discovery reduces drug failure rates, or simply increases the number of candidates that fail.

The second clear use case is in design. For most of its history, drug discovery has been a search: screen enormous libraries of existing molecules and hope one binds. AI has begun to change the task from finding to building. Having learned to predict a protein's shape, it can now design new proteins to order, specifying functions and generating structures to fit. That turns part of discovery from searching a haystack into designing the needle.

The third involves a combination of AI and quantum computing.

Both the pipeline and the design work run into the same wall. They come down to modelling molecules, and molecules are quantum objects. The behaviour of their electrons, which decides whether a drug binds tightly, selectively and safely, is governed by quantum mechanics, and ordinary computers can only approximate it. A quantum computer runs on those same rules, so in principle it could capture the fine detail that classical methods are forced to smooth over.

Researchers are already pairing the two, building quantum versions of the same structure-prediction and molecule-design techniques AI uses, in a field now called quantum machine learning.

Results are promising, but only at the smallest scale, such as folding peptides a few amino acids long, which remains a long way from a working drug. The hardware has turned a real corner, but useful quantum computing is still reckoned to be 15 to 20 years away.

But so far none of this lands in the consulting room. 

What changes in the clinic

So if all these AI benefits happen upstream, does AI actually promise any change in the clinic?

The early wins for AI in the clinic point two ways.

The first could ease the administrative burden. Ambient AI scribes can record patient consultations, drafting notes for the clinician to review and approve After a randomised trial, UW Health rolled the technology out to around 800 clinicians, where it cut documentation time by about 30 minutes a day and eased burnout.

However, a wider study across 1,800 clinicians found a more modest 16 minutes saved per eight hours of care, not including training time needed to use the technology.

The second win relates to clinical decision-making. Randomised trials have used AI to help more patients stay within their target glucose and blood-pressure ranges and to lower post-operative pain. The sharpest case is the heart: doctors built a virtual copy of each patient's heart, simulated where to operate, then ablated exactly there. Eight of ten stayed free of arrhythmia for a year against a usual rate near 60 percent. 

What' is striking is that,neither of these use cases involve replacement. The clinician stays relevant and necessary. The scribe drafts, but the clinician signs. The twin rehearses, but the clinician operates.

The Real Singularity

Perhaps the whole medical AI debate revolves around a single question. What kind of intelligence does medicine actually need?

Pattern recognition at scale belongs to machines. The judgement that unites science and human need, using information such as methylation signatures, biometric data, family histories, psychological states and personal values, belongs to a clinician.

AI is brilliant at discovering patterns across populations. Medicine needs to integrate these discoveries for the benefit of individuals. The clinician will remain the point where these two intelligences meet.

This is the question Prof. Shafi Ahmed, one of the speakers at Next Generation Medicine 2026,  places at the centre of his book, Intelligent: The Evolution of AI Transforming Healthcare.

Prof. Ahmed is no bystander to technology. He streamed the first operations via Google Glass and virtual reality to students in more than a hundred countries, has advised Google and Johnson & Johnson on medicine, and was named among the top fifty innovators in medical AI.

His verdict is optimistic and precise: AI belongs inside the integrative practice, as a tool in the clinician's hands rather than a replacement for them. The doctor of 2030 looks less like a lone diagnostician and more like the integrator who decides which machine outputs to trust, how to combine them with what they learn about a patient, and when to overrule them.

That is the real singularity, and it looks nothing like the one we were warned about: two kinds of intelligence converging in one room, with the human still holding the pen.

The conversation in Dubai

AI earns its place in the clinic only when it is backed by robust evidence, ethical oversight and clear clinical accountability. That raises a more useful question: what are the ethics of AI care, and how might they reshape the economics of the clinic of 2030?

At Next Generation Medicine 2026, the programme looks at how human and artificial intelligence can work together in longevity medicine and anti-ageing care.

Over four days at Atlantis The Royal, clinicians, scientists and innovators from more than twenty countries will present research and debate the hard cases in longevity, regenerative and stem cell medicine. 

Next Generation Medicine 2026.
7 to 10 November. Atlantis The Royal, Dubai.
Get tickets now.