Why Artificial Intelligence has not yet revolutionized healthcare

Artificial intelligence and machine learning are predicted to be part of the next industrial revolution and could help business and industry save billions of dollars by the next decade.

The tech giants Google, Facebook, Apple, IBM and others are applying artificial intelligence to all sorts of data.
Machine learning methods are being used in areas such as translating language almost in real time, and even to identify images of cats on the internet.

So why haven't we seen artificial intelligence used to the same extent in healthcare?
Radiologists still rely on visual inspection of magnetic resonance imaging (MRI) or X-ray scans – although IBM and others are working on this issue – and doctors have no access to AI for guiding and supporting their diagnoses.



The challenges for machine learning

Machine learning technologies have been around for decades, and a relatively recent technique called deep learning keeps pushing the limit of what machines can do. Deep learning networks comprise neuron-like units into hierarchical layers, which can recognize patterns in data.
This is done by iteratively presenting data along with the correct answer to the network until its internal parameters, the weights linking the artificial neurons, are optimized. If the training data capture the variability of the real-world, the network is able to generalize well and provide the correct answer when presented with unseen data.

So, the learning stage requires very large data sets of cases along with the corresponding answers. Millions of records, and billions of computations are needed to update the network parameters, often done on a supercomputer for days or weeks.
Here lie the problems with healthcare: data sets are not yet big enough and the correct answers to be learned are often ambiguous or even unknown.

We're going to need better and bigger data sets

The functions of the human body, its anatomy and variability, are very complex. The complexity is even greater because diseases are often triggered or modulated by genetic background, which is unique to each individual and so hard to be trained on.
Adding to this, specific challenges to medical data exist. These include the difficulty to measure precisely and accurately any biological processes introducing unwanted variations.

Other challenges include the presence of multiple diseases (co-morbidity) in a patient, which can often confound predictions. Lifestyle and environmental factors also play important roles but are seldom available.
The result is that medical data sets need to be extremely large.



This is being addressed across the world with increasingly large research initiatives. Examples include Biobank in the United Kingdom, which aims to scan 100,000 participants.

Others include the Alzheimer's Disease Neuroimaging Initiative (ADNI) in the United States and the Australian Imaging, Biomarkers and Lifestyle Study of Ageing (AIBL), tracking more than a thousand subjects over a decade.
Government initiatives are also emerging such as the American Cancer Moonshot program. The aim is to "build a national cancer data ecosystem" so researchers, clinicians and patients can contribute data with the aim to "facilitate efficient data analysis". Similarly, the Australian Genomics Health Alliance aims at pooling and sharing genomic information.

Eventually the electronic medical record systems that are being deployed across the world should provide extensive high quality data sets. Beyond the expected gain in efficiency, the potential to mine population wide clinical data using machine learning is tremendous. Some companies such as Google are eagerly trying to access those data.

What a machine needs to learn is not obvious

Complex medical decisions are often made by a team of specialists reaching consensus rather than certainty.
Radiologists might disagree slightly when interpreting a scan where blurring and only very subtle features can be observed. Inferring a diagnosis from measurement with errors and when the disease is modulated by unknown genes often relies on implicit know-how and experience rather than explicit facts.

Sometimes the true answer cannot be obtained at all. For example, measuring the size of a structure from a brain MRI cannot be validated, even at autopsy, since post-mortem tissues change in their composition and size after death.



So a machine can learn that a photo contains a cat because users have labelled with certainty thousands of pictures through social media platforms, or told Google how to recognizes doodles.
It is a much more difficult task to measure the size of a brain structure from an MRI because no one knows the answer and only consensus from several experts can be assembled at best, and at a great cost.

Several technologies are emerging to address this issue. Complex mathematical models including probabilities such as Bayesian approaches can learn under uncertainty.

Unsupervised methods can recognize patterns in data without the need for what the actual answers are, albeit with challenging interpretation of the results.

Another approach is transfer learning, whereby a machine can learn from large, different, but relevant, data sets for which the training answers are known.

Medical applications of deep learning have already been very successful. They often come first at competitions during scientific meetings where data sets are made available and the evaluation of submitted results revealed during the conference.
At CSIRO, we have been developing CapAIBL (Computational Analysis of PET from AIBL) to analyze 3-D images of brain positron emission tomography (PET).



Using a database with many scans from healthy individuals and patients with Alzheimer's disease, the method is able to learn pattern characteristics of the disease. It can then identify that signature from unseen individual's scan. The clinical report generated allows doctors to diagnose the disease faster and with more confidence.
In the case (above), CapAIBL technology was applied to amyloid plaque imaging in a patient with Alzheimer's disease. Red indicates higher amyloid deposition in the brain, a sign of Alzheimer's.

The problem with causation

Probably the most challenging issue is about understanding causation. Analyzing retrospective data is prone to learning spurious correlation and missing the underlying cause for diseases or effect of treatments.
Traditionally, randomized clinical trials provide evidence on the superiority of different options, but they don't benefit yet from the potential of artificial intelligence.

New designs such as platform clinical trials might address this in the future, and could pave the way of how machine learning technologies could learn evidence rather than just association. So, large medical data sets are being assembled. New technologies to overcome the lack of certainty are being developed. Novel ways to establish causation are emerging.

This area is moving fast and tremendous potential exists for improving efficiency and health. Indeed, many ventures are trying to capitalize on this. Startups such as Enlitic, large firms such as IBM, or even small businesses such as Resonance Health, are promising to revolutionize health.

Impressive progress is being made but many challenges still exist.

by Olivier Salvado, The Conversation

YOUR INPUT IS MUCH APPRECIATED! LEAVE YOUR COMMENT BELOW.

Why it's dangerous to outsource our critical thinking to computers

It is crucial for a resilient democracy that we better understand how Google and Facebook are changing the way we think interact, and behave

The lack of transparency around the processes of Google’s search engine has been a preoccupation among scholars since the company began. Long before Google expanded into self-driving cars, smartphones and ubiquitous email, the company was being asked to explain the principles and ideologies that determine how it presents information to us. And now, 10 years later, the impact of reckless, subjective and inflammatory misinformation served up on the web is being felt like never before in the digital era.
Google responded to negative coverage this week by reluctantly acknowledging and then removing offensive autosuggest results for certain search results. Type “Jews are” into Google, for example, and until now the site would autofill “Jews are evil” before recommending links to several rightwing anti-Semitic sites.

What follows, the misinformation debacle that was the US general election. When Facebook CEO Mark Zuckerberg addressed the issue, he admitted that structural issues lie at the heart of the problem: the site financially rewards the kind of sensationalism and fake news likely to spread rapidly through the social network regardless on its veracity. The site does not identify bad reporting, or even distinguish fake news from satire.
Facebook is now trying to solve a problem it helped create. Yet instead of using its vast resources to promote media literacy, or encouraging users to think critically and identify potential problems with what they read and share, Facebook is relying on developing algorithmic solutions that can rate the trustworthiness of content.



This approach could have detrimental, long-term social consequences. The scale and power with which Facebook operates means the site would effectively be training users to outsource their judgment to a computerized alternative. And it gives even less opportunity to encourage the kind of 21st-century digital skills – such as reflective judgment about how technology is shaping our beliefs and relationships – that we now see to be perilously lacking.

The engineered environments of Facebook, Google and the rest have increasingly discouraged us from engaging in an intellectually meaningful way. We, the masses, aren’t stupid or lazy when we believe fake news; we’re primed to continue believing what we’re led to believe.

The networked info-media environment that has emerged in the past decade – of which Facebook is an important part – is a space that encourages people to accept what’s presented to them without reflection or deliberation, especially if it appears surrounded by credible information or passed on from someone we trust. There’s a powerful, implicit value in information shared between friends that Facebook exploits, but it accelerates the spread of misinformation as much as it does good content.



Every piece of information appears to be presented and assessed with equal weight, a New York Times article followed by some fake news about the pope, a funny dog video shared by a close friend next to a distressing, unsourced and unverified video of an injured child in some Middle East conflict. We have more information at our disposal than ever before, but we’re paralyzed into passive complacency. We’re being engineered to be passive, programmable people.

In the never-ending stream of comfortable, unchallenging personalized info-attainment there’s little incentive to break off, to triangulate and fact check with reliable and contrary sources. Actively choosing what might need investigating feels like too much effort, and even then a quick Google search of a questionable news story on Facebook may turn up a link to a rehashed version of the same fake story.

The “transaction costs” of leaving the site are high: switching gears is fiddly and takes time, and it’s also far easier to passively accept what you see than to challenge it. Platforms overload us with information and encourage us to feed the machine with easy, speedy clicks. The media feeds our susceptibility to ‘filter bubbles’ and capitalizes on contagious emotions such ‘anger’.

It is crucial for a resilient democracy that we better understand how these powerful, ubiquitous websites are changing the way we think, interact and behave. Democracies don’t simply depend on well-informed citizens – they require citizens to be capable of exerting thoughtful, independent judgment.

This capacity is a mental muscle; only repeated use makes it strong. And when we spend a long time in places that deliberately discourage critical thinking, we lose the opportunity to keep building that skill.



Source: Evans Selinger is a professor of philosophy at Rochester Institute of Technology, and Brett Frishmann is the Microsoft visiting professor of information technology policy at Princeton University and professor of law at Benjamin N Cardozo School of Law. Their forthcoming book Being Human in the 21st Century (Cambridge University Press, 2017) examines whether technology is eroding our humanity, and offers new theoretical tools for dealing with it.

YOUR INPUT IS MUCH APPRECIATED! LEAVE YOUR COMMENT BELOW.

Telepathy and the future

dvances in the field of artificial intelligence are invariably greeted with concern about an imminent robot uprising. Similarly, when we hear about developments in the field of brain-to-brain communication, we imagine any number of outlandish scenarios: perhaps a government marching us unquestioningly into battle via a process of insidious mind control, or an erotic thought we had about a work colleague being unwittingly transmitted to our partner.

When Facebook’s CEO, Mark Zuckerberg, announced this week during one of his regular Q&A sessions that Facebook is working in the field of thought transmission, we found ourselves momentarily transported to a horrific telepathic future. “You’ll just be able to think of something and your friends will immediately be able to experience it too,” he said, as people thought to themselves “under no circumstances do I want anyone to know the dark, unsettling images that flash through my mind on an hourly basis”. We are troubled by that vision. But it’s only a vision.

When similar announcements have slipped out from the University of Washington or Harvard Medical School, people have barely noticed. But the words “Zuckerberg” and “Facebook” seem to prompt disproportionate fury; some people immediately imagined their thoughts directly transposed on to a Facebook wall, while others maintained their long-held position that “Facebook is awful because I say so”. But Zuckerberg’s allowed to dream. He’s allowed to ponder the trend from sharing photos to sharing video, and how that might extend to immersive virtual reality. Facebook’s three AI labs in New York, Menlo Park and Paris are currently working on technology that refines methods of deciding what we want to see and what we want to know. So it’s hardly surprising that the CEO of a social media company would airily predict some kind of all-encompassing sensory experience that, in his words, would enable “richer relationships with people we love and care about”.

But the precise nature of such predictions are so unfathomable that they may as well be outlines for a screenplay. Yes, scientists have managed to facilitate brain-to-brain communication – first between rats, then between human and rat, and then between human and human. Two years ago, scientists at the University of Washington managed to harness the brain signal from one person, and use that signal to magnetically stimulate the motor cortex of another person, inducing them to press the fire button on a computer game. Boom!

Last year an experiment was conducted whereby a computer interpreted someone in France thinking the word “hola”, and sent that information across the internet to India, where it was converted into a series of lights, induced in someone’s brain, that also represented the word “hola”. Yes-no responses have been transmitted between people using flashing strobes and non-invasive brain stimulation. But these experiments, while impressive, require very elaborate setups to crudely stimulate relatively large areas of the brain.

The kind of rich emotional telepathic experience Zuckerberg daydreams about would involve, one presumes, precise manipulation of the 100 billion neurons in the human brain – neurons that have more connections between them than there are atoms in the universe. In addition, brain-to-brain communication would involve the mediation of a computer – but it’s far from proven that a computer could ever capture nuanced thought, let alone things like sentience, consciousness and the way we feel.

Fierce arguments reverberate about the nature of the singularity, the point where artificial intelligence matches our own; some people, like Ray Kurzweil of Google, have a utopian view of man and machine in perfect harmony, while others believe that humans would be lucky to survive it. But it’s all conjecture. Zuckerberg might imagine our fleeting thoughts being transmitted with perfect clarity across Facebook, but a more realistic vision of the future might be of one person thinking of a cow, and someone else saying “are you thinking of a cow?”

During that Q&A, Zuckerberg said something that perhaps warranted more discussion than all the mind-melding stuff. “Our lives improve as our communication tools get better in many ways,” he said. “The increase in the power people have to share is one of the major forces driving the world today.” Again, it’s not surprising to hear this assertion from the head of a social media company but it’s far from proven to be true. Zuckerberg’s vision is one of infinite communication – but that’s not something our brains can cope with. If we eat a lot of food our brain knows we feel full, but in evolutionary terms we’ve never had to develop a feeling of being “full” of communication. The antipathy some people feel about social media might be the beginning of that. And if brain-to-brain communication does one day become a reality, it’s perfectly possible that by then we’ll just think, “no, thanks”.

http://www.theguardian.com/commentisfree/2015/jul/03/telepathy-technology-mark-zuckerberg-facebook-dystopian-future?CMP=share_btn_link
Photograph: Justin Sullivan/Getty Images