The machines are here, and they're smarter than ever. Scientists have proven that machine learning and artificial intelligence (AI) can perform some tasks better than humans. Where health and neuroscience are concerned, one doctor says this is "a good thing."
By using advanced scanners and AI, a team of scientists from research labs all across the United States created the Human Connectome Project (HCP), the most detailed map of the human brain's circuitry. The AI programs were able to learn 100 new regions of the brain, while monitoring data from 210 adult test subjects.
"AI has been used in particular to help us understand images that come off scans and scanners because these images are really rich in terms of the amount of information that they contain," says Dr. Arthur Toga, a neuroscientist at USC's Laboratory of Neuroimaging and one of the scientists involved in the Human Connectome Project. "Using algorithmic approaches to extract information from these data sets has become a tool that's used quite often."
Toga says the approach is comparable to using a face recognition algorithm or military application to find targets in a potentially dangerous setting. During the brain imaging process, the AI system is asked to identify features that it can be taught using prior examples. The algorithm is then able to use a set of strategies to find these features in a training set that can consist of hundreds or thousands of examples. Once the training set has proved successful, the next step is to take what the algorithm has learned and apply it to an experimental set, in order to prove and validate the training.
The question is, could this project be done without AI? Toga says yes, but not as well.
How can AI be used for visual diagnosis?
A similar study took place at Stanford University where researchers discovered that a machine learning algorithm can identify lung cancer tissue more accurately than epidemiologists.
This is happening because the juxtaposition between computers and humans is becoming increasingly tight, and soon the distinction between the two will blur.
What are the benefits of combining artificial intelligence and human intelligence?
Justin Lyon, CEO of the predictive performance platform Simudyne, believes that artificial intelligence needs to merge more with human intelligence. He says the main benefit of such a convergence is speed.
"A group of humans coupled with an algorithm can beat our best AIs today," says Lyon. "We can make computers move so quickly that they can do 40,000 things in the time it takes for you to blink your eye. The speed is incomprehensible for humans, and is now being applied in its learning."
With regard to fears around AI and machines replacing humans or making errors, Lyon said: "I know some people get scared by that, but I'm on the opposite side end. I'm ecstatically optimistic about that because our species is facing absolutely enormous challenges, and we don't have the time as humans to solve them. But we're lucky that just in time we've created computers. It's only with these computers that we have any chance at solving these problems."
Accuracy and speed aren't the only positives of AI and machine learning in the health sector. The technology is also lowering costs. According to the Portland-Oregon based startup Bright.md, its automated system, or "AI nurse," cuts down online doctors appointments from 20 minutes to 90 seconds, and reduces the cost of providing services by 80%.
This sounds like a valuable combination.