Machine learning is solving real problems, from saving energy to combatting blindness


Millions of people go blind every year from a preventable eye disease caused by diabetes.

Called diabetic retinopathy, it is devastating to its sufferers and takes an ophthalmologist to diagnose. This is a particular problem in a country like India, which has a huge population and a high prevalence of diabetes but not enough ophthalmologists.

“India is set to emerge as the diabetic capital of the world. According to the WHO, 31.7m people were affected by diabetes mellitus in India in the year 2000,” says a 2016 study in the Indian Journal of Ophthalmology.

“This figure is estimated to rise to 79.4m by 2030, the largest number in any nation in the world. Almost two-thirds of all type-2 and almost all type-1 diabetics are expected to develop diabetic retinopathy.”

What if you could teach a computer, through machine learning, to recognise the symptoms and make a diagnosis of diabetic retinopathy, much like researchers at Google have done?

Better yet, what if the artificial intelligence (AI) computer that is learning how to do this diagnosis were already an expert at identifying objects for Google’s photographic recognition software?

“We were able to take something core to Google — classifying cats and dogs and faces — and apply it to another sort of problem,” says Lily Peng, the Google lead running the project, who is also a physician and biomedical engineer.

Her paper in the Journal of the American Medical Association, published last November, reported that AI could recognise diabetic retinopathy at the same pace as an ophthalmologist. Given the superior processing power, better recognition software and network effect of a larger database of previously scanned images, you can expect that efficiency to increase.

This project is one of the best examples of the nascent field of AI, which is seen as the next evolution not only of computing but of work. Technically, it’s better to think of it as machine learning, because these are still process-orientated computers which are performing and learning tasks. The word intelligence implies so much more and no computer has evolved to that point yet.

Training computers to think like humans has emerged as a cutting-edge field of endeavour. To speed up the pace Google has built machine-learning software to build machine-learning software, called AutoML.

“Today these are handcrafted by machine-learning scientists and literally only a few thousand scientists around the world can do this,” Google CEO Sundar Pichai said recently.

Google has changed its mission from the “mobile first” mantra espoused by former CEO Eric Schmidt to Pichai’s “AI first” as it shifts to focus on this new major trend.

With the new technology, the eyesight of millions could be saved through early detection of disease. It’s progress we can be proud of.

This column first appeared in Financial Mail


About Author

Toby Shapshak is editor-in-chief and publisher of Stuff, a Forbes contributor and a Financial Mail columnist. He has been writing about technology and the internet for 20 years and his TED Global talk on innovation in Africa has over 1,5-million views. He has written about Africa's tech and start-up ecosystem for Forbes, CNN and The Guardian in London. He was named in GQ's top 30 men in media and the Mail & Guardian newspaper's influential young South Africans. He has been featured in the New York Times. GQ said he "has become the most high-profile technology journalist in the country" while the M&G wrote: "Toby Shapshak is all things tech... he reigns supreme as the major talking head for everything and anything tech."

Leave A Reply