You’ll probably know IBM’s supercomputer, Watson, from its 2011 appearance on Jeopardy. Up against two of the US quiz show’s longest-running and highest-earning contestants, Watson clinched a $1m prize after answering a series of quick-fire general knowledge questions. It wasn’t a close call either at the final score, Watson’s total was $31,547 ahead of its rivals’ combined. But in the five years since, IBM’s supercomputer has been working towards another goal, one far more lucrative than the Jeopardy! jackpot: upgrading healthcare.
Roughly every three years, the amount of medical data on the planet doubles in size. By 2020, it is expected to double every 73 days. But you’d be wrong in thinking this exponential growth is being matched by progress in the field. Our vast pool of expertise research, clinical trials, medical records is a mess. It’s siloed, it’s unstructured and, crucially, it’s polyglottic.
This is where AI systems like Watson come in. These so-called ‘cognitive computers’ can find patterns in large swathes of unstructured data. But not only that, their ability to learn means they can also navigate languages and all their subtle linguistic cues. Essentially, these systems promise to make the medical profession – and everything in it – Google-able.
Say, for example, a patient is wondering whether to go to hospital after finding an unusual rash on their abdomen. They can dial 111 to use the NHS hotline, or – if Dr Ali Parsa has anything to do with it – they’ll use his AI triage service, which debuted in April. Here, in a matter of seconds, their symptoms are checked against a database of more than 100 million valuations to suggest next steps.
“We’ve actually shown in our tests that the machine is more accurate than a nurse with 30 years experience,” said Parsa.
In around two-thirds of cases, the service will recommend a trip to the pharmacist or an over-the-counter drug like Aspirin.
Using the same methods as Parsa’s company, Babylon Health, AI systems can also help doctors – some of the most time-poor people on the planet with their everyday work. Using the rash example, a system could sift through data like the patient’s family history and new drug releases to present a doctor with a crowd-sourced selection of treatments, ranked by their likelihood of working.
Image parsing
And it’s not only text that AI systems can parse.
Back in October, startup Enlitic announced a partnership with Australian radiology service provider Capitol Health to access its large archive of CT scans, MRI scans and X-Rays. Now, Enlitic’s deep learning software is parsing these images to detect things like early-stage lung nodules and tiny wrist fractures.
These are impossibly tiny to us humans, a fracture can measure four pixels across in an X-Ray image 4,000 pixels wide. So far, Enlitic claims it is between 50% and 70% more accurate than a radiologist working alone.
But aside from acting as an unconventional diagnostician – a la Dr Gregory House – AI can help save lives in less dramatic ways. Take, for example, the statistic from the World Health Organisation that around 50% of drugs are not taken as prescribed. UK homes alone contain around £90m-worth of unused prescription medicines.
AI apps can monitor patients’ biometric data to see they are on track with their treatment plan. Should that data change, their doctor can be alerted. Startups like Babylon are also taking this idea one step further, by preventing sickness ahead of time with a predictive analytics engine. With wearable devices, these virtual personal assistants could keep you on track with meal plans, exercises and taking your medication on time, and in the right dosage. This is already happening with glucose monitor Medtronic, which has partnered with IBM’s Watson Health to alert diabetics about changes to their blood sugar levels.
“Our job as entrepreneurs is to imagine a future world and build it. Your body will become like your car – it just won’t break down anymore,” said Parsa. By cutting out risky habits in people’s lifestyles, Parsa says preventable diseases will reduce. According to him, this could lead to extended life spans, and even set us on the path to escaping death for good.
Timeliness
Even if we’re fit and healthy, we’ll still need to rely on drugs to protect us from infectious diseases if we are to live longer lives. But as the reports on the rise of drug-resistant ‘superbugs’ show, our system is reliant on decades-old developments. For example, the most common ‘last resort’ antibiotic, Colistin, was first discovered in 1949.
“I used to work in big pharma, and the way those companies are set up actually makes it much less likely that they will be able to find the unexpected or have in-depth knowledge across a range of pathways and diseases,” says Jackie Hunter, the CEO of Stratified Medical. Following a £40m raise, her company is using AI to rethink pharmaceutical R&D.
Typically, it takes between 10 and 15 years to find a new drug. However, Stratified Medical is hoping to reduce this timespan dramatically. Hunter and her team of biomedical scientists are using deep learning techniques to hunt for patterns between diseases and chemical compounds in over 20 million documents – performing calculations in seconds that would have previously taken a lifetime of work.
It has already sold some proprietary targets to a drug company, but Stratified Medical isn’t just doing heavy-lifting around new compounds, it’s looking to reposition those that are left gathering dust on the shelf: finding new uses for the 95% of compounds that fail to make it to market from clinical trials.
If the stars align and Stratified is able to match a compound to another drug company’s molecule it could deliver a product in just two or three years.
“We’re more agile and nimble and also not hampered by tradition lines of reporting or division,” Hunter added.
Market research firm Frost & Sullivan forecasts that revenues of AI healthcare systems will reach $6,662bn by 2021, with IBM accounting for 45% of the market share.
“AI has become a fashionable world. Previously it was the social network and before that the on-demand economy, now it’s AI,” says Parsa. “Just yesterday, I hired a guy. There are only around 120 people in the world who can do what he does.”
Challenges
One of the biggest challenges for the UK’s AI community right now is attracting and retaining top talent. Ross King, a professor of machine intelligence at the University of Manchester’s School of Computer Science, agreed with Parsa: “The best people in machine learning and artificial intelligence tend to be bought up by Google and Facebook, essentially doing advertising rather than something of societal value.”
Another issue is the perception of the technology. The threat of machines wreaking havoc on our world and displacing the workforce looms large in the minds of consumers. A YouGov survey commissioned by the British Science Association found 36% of respondents thought AI presented a “threat to the long term survival of humanity”, while 53% would not trust a robot to perform surgery.
But Ben Taylor, CEO of startup Rainbird, an AI startup that helps experts model what they know, thinks consumers deserve more credit.
“The way we define AI is any system that performs some function considered uniquely human,” he said, adding: “Sat Nav would have been scary a decade ago, but you [grow to] accept it and move on.”
While European startups don’t raise the kind of eye-watering amounts now common in the Valley, the region has one big advantage over the US.
“Here, we have a healthcare system that is globally recognised. Whereas the US healthcare system is highly siloed, and you can’t fix a silo with a silo,” said Parsa.
Another analyst, Bruce Daley, from Tracita, told Tech City News that, in the United States, the return on investment in AI is mostly negative and is likely to remain so for some time, with other areas of the world likely to adopt the technology much sooner.
Indeed, AI platforms can help small teams perform evidence-driven care on the cheap. Just as mobile networks ‘leapfrogged’ the need for landlines, maybe AI can leapfrog hospitals and support staff to give everyone – even those in remote areas – a doctor in their pocket.
So where next? Perhaps the biggest promise of AI isn’t replacing our top doctors – some skills may always be beyond robots – but bringing their expertise to the rest of the world. If successful, AI systems could take medicine’s ugly secret and harness it to democratise access for good.
Date: July 18, 2016