Artificial intelligence isn’t just about machines beating humans at games like chess and go, and destroying the world in dystopian Terminator-style nightmares. AI is also making great strides in health care.
While the very idea fills some people with horror (including a few doctors who fear losing their jobs and golf club memberships), this technology has the potential to deliver world class diagnostic tools to patients who could never afford to see a specialist.
The cool thing about machines is that while they make mistakes, just like us, they also learn from their mistakes, and all at once, if they’re networked. They’re able to digest as much data as is thrown at them, the more the better, without ever getting bored and going for donuts.
In the UK, a database of 2.5 million anonymous mammogram images has already made it possible for machine learning programs to be better at detecting breast cancers than human radiologists working alone, while also reducing unnecessarily alarming false positive results.
In detecting skin cancers, there are now cheap apps which allow anyone with a smartphone to upload images of worrying skin conditions and access the serious processing power of remote AI to decide whether more action needs to be taken. Results are impressive and becoming more so.
Some companies are already claiming better results than dermatologists.
And it’s not just cancer. Artificial intelligence can speed diagnosis and avoid mis-diagnosis for conditions related to cardiovascular disease and diabetes.
By combining machine learning with EEG recordings, major new advances beckon for conditions including Alzheimers, Parkinsons, stroke, chronic pain, and drug-resistant epilepsy.
As with all things digital, the quality of the outputs depends on the quality of the inputs, but with more advanced forms of AI, such as deep learning, the computers don’t need to be led by the hand every step of the way from evidence to conclusions, but are able to make leaps of their own, and potentially much faster than human experts.
Deep learning technology is also being used to invent new drugs, and do genetic research.
Computer-aided distillation of existing written medical knowledge is another new frontier, leading to connections unfound by humans. Sadly there isn’t yet an app that can translate your doctor’s handwriting.
For all the potential benefits, there are clear risks associated with these technologies, particularly around patient privacy. If AI health is merged with bio-surveillance, as has already been proposed in some countries, we could end up with a frightening scenario, especially for people with compromised health situations.
Government health organisations getting into bed with big data corporations, like Google, present obvious risks, as the reaction to the new COVIDSafe app has shown.
Intelligent machines could also give cost-cutters an excuse to sack medical staff who have additional knowledge and expertise that machines are incapable of. Machines are probably not capable of care, in the human sense, not yet anyway.
Hopefully human medical experts and machines will be able to work together in the future, each playing to their particular strengths. Whatever happens with cyber-medicine, a brave new world for health care beckons.