Computers have for a long time been good at some things, like doing math, but bad at other things we humans are good at.
Marcel Salathé is director of the Digital Epidemiology Lab at EPFL
Computers have for a long time been good at some things, like doing math, but bad at other things we humans are good at. For example, just 10 years ago, it was practically impossible for a computer to look at a few images of different animals, and correctly classify the cat as cat, the dog as dog, etc. – something a two year old can do easily. But research in artificial intelligence has made such tremendous progress in the past few years that computers can now do this just as well as humans, or even better. And they can do it millions of times faster.
We can now build software - based on ideas about how our own brains work - that can “learn” what a cat looks like. And of course, such learning is not limited to images of cats, but can be extended to any situation imaginable. In a recent study for example, a computer learned to identify the correct type of skin cancer based on an image alone - and ended up being better at it than a dermatologist at Stanford Hospital.
Remarkably, building such software is very easy, not at least thanks to the open source movement. The large digital companies of this world are now openly releasing the software that powers artificial intelligence, accelerating the rapid progress in this field even more.
Artificial intelligence thrives on data – and we are in the age of big data. We collect, both individually and institutionally, enormous amounts of data on everything we do. From our DNA up to how many steps we take per day, it costs almost nothing to collect any kind of data you’d want to have about yourself. And when hundreds of millions of people do this, massive amounts of data are generated that artificial intelligence can learn from. But who owns this data?
Artificial intelligence is here to stay. The search engine you use is based on artificial intelligence. The map you use to get from A to B as fast as possible is based on artificial intelligence. Markets move based on decisions made by artificial intelligence. One day soon, many medical diagnoses will be based on artificial intelligence. Whoever controls the data, controls the artificial intelligence. Rather than having this control concentrated in a few commercial players, I hope we can create a situation where all of us – medical professionals, researches, corporations, patients – are in control of this new technology. For this
to happen, all stakeholders need to work together
to find ways to open medical data to the wider
community. Otherwise, our future medical system risks being controlled by a select few who have
access to the data. ⁄
Marcel Salathé is an associate professor at EPFL, where he heads the Digital Epidemiology Lab. He is the author of Nature, in Code and a co-founder of PlantVillage, a knowledge sharing platform on plant diseases.