In the 1966 film Fantastic Voyage, a submarine crew is shrunken to microscopic size and injected into the bloodstream of a defector to remove a blood clot from his brain. Critics agreed that it was an entertaining movie but that the impossible premise took some swallowing.
Last month John McNamara, a leading IT specialist at IBM’s research and development laboratory in Hursley, Hampshire, suggested to the House of Lords Artificial Intelligence Committee that within 20 years ‘We may see AI nano-machines being injected into our bodies. These will provide huge medical benefits, such as being able to repair damage to cells, muscles and bones — perhaps even augment them.’
We now read daily in newspapers the sort of stories once found only in the pages of the most fantastical science fiction novels. ‘AI becomes world’s best Go player in just three days’; ‘Self-driving cars could run on unlit roads to conserve energy’; ‘Robot behaviour is creeping beyond our control.’ These were all recent headlines in the Guardian, the Times and the Financial Times.
McNamara, and many others, believe we are entering what they call the Second Machine Age. The Industrial Revolution heralded the advent of the First Machine Age, in which machines reproduced human physical labour. The Second Machine Age will see human mental capabilities reproduced and bettered and it will have a huge impact on medicine and healthcare.
Britain, incidentally, is a world leader in AI. That ‘world’s best Go player’, which absorbed 3,000 years of knowledge of the board game in just three days and then improved upon it, was created by London-based DeepMind, which acquired by Google in 2014 for $400 million (about £309 million).
DeepMind’s health division is currently working with the NHS foundation trusts of University College London Hospitals and Moorfields Eye Hospital on a research project to analyse medical scans.
Twitter and Amazon have, in recent years, snapped up British AI outfits – something that the government is keen to point out. Culture Secretary Karen Bradley has said: ‘I want the UK to lead the way in artificial intelligence. It has the potential to improve our everyday lives — from healthcare to robots that perform dangerous tasks.’
The AI revolution is being driven by neural networks. Broadly speaking, a conventional computer programme is a detailed and precise set of instructions for accomplishing a particular task. A neural network is a collection of linked digital processing units roughly modelled on the human brain. It ‘learns’ — becomes more efficient — by being presented with examples.
To teach a neural network to identify a beech tree, you feed through it digitised images of beeches in all shapes and sizes, essentially saying ‘Look! Beech!’, and pictures of other types of tree, saying ‘Not beech!’. Once ‘trained’, the network will correctly identify new hitherto unseen images.
To accomplish the same task in a conventional programme of yesteryear would involve writing screeds of code attempting precisely to define the image of a beech tree as distinct from an ash, sycamore and so on.
Pietro Valdastri is professor of robotics and autonomous systems at Leeds University and director of the Science and Technologies Of Robotics in Medicine Lab at Vanderbilt University in Nashville. ‘We are witnessing an acceleration in AI-robotics mainly driven by the automotive industry with self-driving cars,’ he says. ‘Similarly to self-driving cars that are able to park without human intervention, medical robots of the next generation will be able to perform autonomous sub-tasks.
‘They will be able to complete autonomous sutures and other time-consuming repetitive tasks under the supervision of the surgeon, achieving a higher reliability than the human operator.
‘AI will help in the visual diagnosis of diseases by recognising lesions from images where clues are not evident to the human eye.
‘Artificial limbs and exoskeletons will recognise the type of context — for example, walking up steps, running, dancing — and will adapt autonomously, providing a smoother experience to the user.
‘And one of the most exciting innovations is the possibility of having advanced diagnostic tools for procedures such as gastroscopy or colonoscopy (for gastric or colorectal cancer screening and diagnosis) that, thanks to soft robotics and AI, will be easy to use and so ultra-low cost that they will be available in GP practices.’
Cancer is still seen by many as the condition that medicine must strive to conquer. Mortality rates for some cancers such as testicular have plummeted. But for others, such as pancreatic, brain and oesophageal, little progress has been made in recent decades.
Dr Áine McCarthy, of Cancer Research UK, says: ‘The point we are at in our understanding of cancer, and with the level of AI technology we have reached, means we expect some of the potential of using AI in cancer research to start to be realised soon.
‘One of the things researchers are looking at is whether AI could help improve cancer diagnosis. Could it help doctors to spot cancer more easily in its very early stages, which can be difficult to do using current conventional means? Where AI and machine learning could help with early diagnosis is pattern recognition.
‘Pathologists are incredible at what they do but if we were able to use AI and machine learning to recognise patterns on scans or tissue samples as well, it could be used alongside these specialists to improve and potentially speed up cancer diagnosis.’
Eleni Vasilaki, professor of computational neuroscience at Sheffield University, believes that over the next decade AI’s most significant impact will be in the field of personalised medicine.
‘Not all people respond to treatments in the same way,’ she says. ‘For instance, there are genetic differences that can affect the effectiveness of drugs. AI can help sort patients into different groups and then identify the most appropriate and effective treatment.’
Professor Vasilaki is, like many experts, cautious about the Fantastic Voyage-like vision of IBM’s John McNamara, who posits miniature robots at work inside our bodies. ‘I don’t think we are anywhere near this stage,’ she says. ‘The research vision is there, but I would not anticipate this happening in the near future.’
But then, 50 years ago, the very idea was nothing more than far-fetched speculative sci-fi. Now, the reality of it may lie just beyond the near future and that is getting closer every day.