In the tenth anniversary year of the Big Bang in AI we will see better drug discovery processes, better diagnostics, and better understanding of human biology.

Tenth anniversary of AI’s Big Bang

Next year will see the tenth anniversary of the Big Bang in artificial intelligence (AI). In September 2012, a team led by Geoff Hinton won an annual image recognition competition called ImageNet, using a type of silicon chip called graphics processing units (GPUs) and a type of algorithm called convolutional neural nets (CNNs). Hinton’s team was not the first to use these techniques, but the margin of their victory was startling. It ignited enormous excitement about AI, and in particular, about subsets of AI called machine learning and deep learning. The wave of research and investment this set off has given us machines with superhuman abilities in face recognition, game playing, and language generation and translation models.

The best AI systems today are still a long way from the human-level general intelligence (known as artificial general intelligence, or AGI) which would lead quickly to superintelligence. The Terminator is not lurking just over the time horizon. But modern AI is capable of remarkable achievements, and we are well into the process of deploying this tool across all sectors of the economy – including healthcare.

What new advances in healthcare will AI bring us during the tenth anniversary of its Big Bang? The safest prediction is that we will see more of what we saw in 2021 and previous years: better understanding of human biology, better drug discovery processes, and better diagnostic techniques.


Human biology is a complex affair, and proteins are at the centre of it. Proteins are chains of amino acids. There are billions of them in each human cell, and they catalyse and guide many of the functions of those cells. The way a protein works is determined by the way it folds in three dimensions, and this is fantastically complicated: understanding it is a grand challenge of biology. In July 2021, DeepMind, a subsidiary of Google and a leading AI research organisation, published the details of AlphaFold 2, a programme which predicts how proteins fold with far greater accuracy and speed than was possible before.

DeepMind is hopeful of making further advances with AlphaFold. Combining its AI-driven computational analysis of proteins with improving scanning techniques, like cryogenic electron microscopy (cryo-EM), will deepen our understanding of the detailed workings of human biology in ways that will improve healthcare enormously. The authoritative State of AI Report predicted that DeepMind will announce a major research breakthrough in the physical sciences during 2022.

Eroom’s Law

The complexity of human biology makes the development of drugs extremely challenging. As long ago as the 1980s, scientists noticed that the cost of pharmaceutical R&D was doubling every decade or so. This was later described as Eroom’s Law, since it was the reverse of the exponential improvement in computing known as Moore’s Law.

AI is starting to help. During 2021, the UK startup Exscientia and the Hong Kong-based Insilico Medicine both used AI to develop new drugs to the point of human clinical trials. Exscientia is tackling obsessive-compulsive disorder (OCD), and Insilico is addressing various forms of fibrosis, a damaging type of tissue scarring. By using AI, these firms have dramatically reduced the time and expense involved in drug development. The successes of Exscientia and Insilico are very likely to be replicated next year both by themselves, and by other biotech firms.

As with drug development, AI helps speed up the creation of vaccines. Moderna and BioNTech, which developed the two leading mRNA (messenger RNA) vaccines against Covid-19, were able to get their products into arms much quicker thanks to the efficiency gains from using AI. Sadly, Covid-19 is not done with us yet, so 2022 will doubtless see more AI-assisted vaccine development.

Radiologists and the Roadrunner

Along with drug and vaccine development, medical diagnosis is another domain where AI can contribute. Back in 2016, Geoff Hinton compared radiologists to the coyote in the Roadrunner cartoons. He said they had run off the edge of a cliff but they had not yet looked down. It was obvious, he said, that within five years or so, deep learning systems would be better at reading scans than humans, and therefore no more humans should be trained as radiologists.

Hinton was not the first AI researcher to fall foul of Amara’s Law, which states that we often over-estimate what a new technology can achieve in the short term, while under-estimating what it will achieve in the long term. Sadly it is rarely clear in advance how long the short and the long term are. But the direction of travel is clear. For instance, “dry” age-related macular degeneration (AMD) is hard to detect in the early stages, and can lead to blindness if left untreated. Moorfields Eye Hospital in London has now developed a computer vision system that can detect it. 2022 will doubtless see more innovations like this.

Diagnosis is not limited to hospitals. Many of us carry diagnostic equipment around with us, in the form of smartphones, watches and rings. These devices monitor our heart rate, our sleep, our temperature, our breathing, and our activity levels. The Apple Watch is one of the most popular, and a new version is released each year. Apple is known to be working on offering a glucose test through its watch, and 2022 might well be the year that they launch it.

We have seen how the tenth anniversary year of AI’s Big Bang will see advances in the understanding of biology, in drug development, and in diagnosis. In addition, there are broad trends in the AI sector generally which will impact healthcare in ways that are hard to predict.

AI systems get both bigger and smaller

One of those trends is contradictory. The largest AI models are enormous, and getting bigger. When OpenAI launched GPT-3 in 2020, it was by far the largest language processing model in the world, with 175 billion parameters. (Parameters are analogous to the synapses in a human brain.) Today there are models with 10 trillion parameters. These models are trained on gigantic amounts of data, using colossal amounts of compute power. These resources, and the skilled researchers able to wrangle them, are available only to the largest companies and the elite universities which those companies sponsor. Some argue this means that AI is being de-democratised.

But at the same time, the low-code, no-code movement is creating software modules that less specialised developers – and even non-programmers – can use to build AI systems of use to their own organisations. The leading AI researcher Andrew Ng argues that sometimes, data integrity can beat data quantity, so that organisations with relatively small lakes of proprietary data can compete on equal terms with much larger entities.


Another trend is increasing automation. As AI becomes more capable, it takes over more functions previously carried out by humans. This makes processes more efficient, and healthcare is no exception: there is plenty of scope to improve workflow in processing patient records, arranging appointments, and so on.

Finally, one trend that is specific to healthcare is that AI is helping both scientists and medical practitioners to see aging as a disease in its own right, rather than as an inevitable component of the human condition which must be tolerated forever. In the last three hundred years we have made great progress against the diseases of youth and of infection, so that most people now die of the diseases caused by aging: cancer, heart disease, and dementia. A clinical trial called TAME (Treating Aging with Metformin) is the first FDA-approved trial of an explicitly anti-aging drug.

Financial institutions are starting to allocate significant amounts of money to tackling aging. These amounts are still small compared to the overall resources available to pharmaceutical and medical companies, but they are growing quickly. A possibly even more significant source of funds for anti-aging research is philanthropists. In October, a company called Altos Labs raised around $300m from billionaires including Jeff Bezos. Many will be cynical about their motives, but technologies which extend lives will benefit everyone.

Extending both healthspan and lifespan

We started this article with some safe predictions. Let’s end it with a more adventurous one. Barring unforeseen disasters, 2022 will surely see AI helping scientists and clinicians to improve our health span. Maybe it will also bring hope that they can improve our lifespan too. The number of centenarians in the world is booming, but at the moment, no-one seems to make it past about 120. Most people assume without much thought that this number is a permanent hard stop. Next year, developments in the funding and the science of longevity research may see that assumption challenged.

Related Posts