Monday, August 28, 2017

Inside AI: Meeting Robots That Can Detect Alzheimer's and Depression


Just 45 seconds in the company of scientist Frank Rudzicz and his machines is all it takes to determine whether or not you are suffering from Alzheimer’s disease.

In that time, the complex Artificial Intelligence (AI) algorithms that the 37-year-old and his team have developed are able to pick apart your voice and predict the severity of the disease to an accuracy of around 82 per cent (and rising).

First, there is your actual use of language. Alzheimer’s sufferers tend to leave longer pauses between words, prefer pronouns to nouns (for example, saying “she” rather than a person’s name) and give more simplistic descriptions, such as a “car” rather than the model or make.

Then there is what Rudzicz calls the “jittter and shimmer” of speech; variations in frequency and amplitude. “These are very difficult for the human ear to pick up but the computer is objective and completely quantifiable,” he says.

Rudzicz is speaking from the boardroom of WinterLight Labs, the company he co-founded on the upper floors of the West Tower of Toronto’s Mars Discovery District: a cluster of shiny downtown buildings run by a public-private partnership where some of the most ground-breaking AI research in the world is taking place, and from where the Telegraph is reporting for a three-part series on the technologies already changing our lives.

Much has been made in recent days of the world-destroying potential of AI. Last Monday, the founders of more than 160 companies, including Elon Musk of Tesla and British tech entrepreneur Mustefa Suleyman, signed an open letter to the UN warning without urgent action lethal autonomous weapons will create a “third revolution” in warfare similar to that which followed the invention of gunpowder and the atomic bomb.

Despite the very real fears that such technologies could be our undoing, it is also hoped if managed properly they will be of huge benefit to society – nowhere more so than in healthcare where a revolution is already underway.

According to a recent industry projection by the market research company Frost and Sullivan, in 2021 AI in health will be worth £5bn globally, representing a 40 per cent growth on today. 

In April, British digital healthcare company Babylon raised nearly £50m to build an AI doctor that can diagnose illnesses without help from a human.

Frank Rudzicz is working towards something similar. As well as his 45-second test which studies 400 different variables of speech, he has built a robot named Ludwig, two foot-tall and possessing the appearance of a ventriloquist’s dummy.

Ludwig runs on so-called machine learning algorithms which recognise data and make predictions - similar to how Amazon might suggest a new book and Netflix a must-watch box-set. For Ludwig, these algorithms enable him to engage patients in conversation and assess speech patterns to determine their health.

As well as testing for memory and speech impairment, such technology can even predict emotions – and whether or not a patient is at risk of an imminent bout of anxiety or depression.

AI timeline

Rudzicz, who is also an assistant professor in computer science at the University of Toronto, admits there are complex regulatory issues around the extent to which AI machines should be used to diagnose patients.

Currently, his models are being piloted in the largest network of retirement homes in North America, and among elderly patients in Edinburgh and Nice, to collect data and train the machines to understand different languages and accents.

At present, they are only being used only to map cognitive decline within existing patients rather than actually diagnosis new ones.

“We have always been careful to position this as an assessment aid rather than straight diagnosis,” Rudzicz says. “One of the main risks I see with AI in healthcare is people can put a lot of faith into it and discount other sources of evidence.”

How long such restraint continues, though, remains uncertain. Already we rely on AI algorithms contained within our smartphone to map many of our vital statistics: blood pressure, heart rate, sleep quality and fertility.

All the experts predict that in the coming years this ceding of our biological data to machines will rise exponentially to the point where each of us carries around what is, in essence, our own portable GP.

According to Android Dreams, the new book written by the eminent Australian artificial intelligence professor Toby Walsh, smartphones may also take selfies to identify suspect melanomas and monitor the health of eyes. AI-equipped toilets, meanwhile, will unprompted analyse samples of urine and stool and alert us to anything amiss.

In his book, Professor Walsh also offers another prediction: that by 2050, many of us will have had our genes sequenced making it far easier to identify and treat genetic disorders which presently affect some 350m people worldwide.

In a different building in Toronto’s Mars Discovery District, another pioneer in the field of artificial intelligence is working on that exact problem. The aim of Brendan Frey’s work is simple. “We want to change medicine,” he says.

Brendan Frey of the Deep Genomics Credit: Julian Simmonds

The 48-year-old, who is a professor at the University of Toronto and chief executive of the AI health research company Deep Genomics, which he started in 2014, has painful personal experience of the current knowledge gap in genetic disease.

In 2002, he and his wife were told their third child with whom she was pregnant at the time could be suffering from an (unnamed) genetic disorder.

“We were told it could be nothing, or it could be a disaster,” Frey recalls. “It was very difficult to deal with, and we ended up terminating the pregnancy.”

At the time, Frey was on the technical advisory board of Microsoft working on speech recognition. Following the death of his unborn child, he decided to leave and begin focusing on developing the technology that could cure genetic disorders.

Leaning back against a white chalkboard scrawled with impregnable equations and wearing an AC/DC Highway to Hell T-shirt, Frey attempts to explain how his work will unravel the mysteries of the human genome and help both predict and eventually treat diseases such as spinal muscular atrophy and Duchenne muscular dystrophy.  

A man walks past a digital representation of the human genome Credit: Getty

In the Nineties, Frey worked on early AI machine-learning algorithms with the so-called “Godfather of AI”, the British scientist Professor Geoffrey Hinton. Machine learning works by teaching the computer through layers of code, enabling it to build up a pattern of understanding which it can then apply itself to a particular problem – in this instance, mapping the genome.

“The basic fact is no human or group of humans will ever understand how the genome works,” Frey says. “We have an exponentially growing set of data to allow us to peer into cells and read out what is changing. There is only one solution: artificial intelligence. It’s the best technology we have in our systems to understand complex data.”

Certainly Frey’s work is exciting enough to be attracting a lot of venture capital money. He says some £3m was initially raised to start the company and now he is looking to raise a further £9.5m in the coming months, with his 20-strong staff expected to soon double.

According to Frey, pathologists in different laboratories will disagree over a particular genetic mutation up to 50 per cent of the time. In the new machine age he is working to bring about, he says such contrarian advice will be eradicated.

This gulf between robot intelligence and human uncertainty has unsurprisingly fuelled growing talk of AI replacing doctors, radiologists and laboratory technicians.

Frey insists “we will always need humans to address the outliers”. But how long before we will be making an appointment with a machine? 

The robot, not the doctor, will be seeing you soon enough.



All Hypnosis Feeds

via Psychology Headlines Around the World http://ift.tt/TRVxbY

No comments:

Post a Comment