Here are your top stories in AI and Data Science from around the world...
Deep Learning used to detect anorexia in social media posts
Social media posts are filled with revelatory and often painfully personal information about the individual writing them. However, when these posts are seen in isolation, they may camouflage patterns that reveal the potential presence of serious mental health issues. These can range from suicidal thoughts to post-traumatic stress to eating disorders.
Elham Mohammadi (MSc 19) and Hessam Amini, who is working toward his PhD, looked at posts by Reddit users labelled anorexic or non-anorexic and a collection of their posts recorded chronologically. Using deep learning algorithms, the researchers looked for patterns of linguistic features in the posts that would identify the individual as being anorexic or at risk of anorexia.
The authors were far more interested in early detection of mental health issues than detecting harmful behaviour after the fact – something that has so far been under-researched, they say.
Stevie the Robot the latest face on Time Magazine
A team of Irish designers are developing a prototype which can attend to the social care needs of older people. Trundling around on three wheels, Stevie’s face comprises a screen which can host pictorial prompters for those with hearing impairments, and may even be used to independently contact emergency services if the user becomes unresponsive.
Such applications may be useful for providing companionship among an often isolated cohort as the population ages, and can help to assuage more acute fears such as the risk of falls, medication errors or carbon monoxide poisoning according to project leader Prof Conor McGinn.
Machine Learning Predicts Thyroid Cancer Risk
A machine learning algorithm was able to detect malignant thyroid nodules in ultrasounds with 77 per cent accuracy, offering a fast and inexpensive way to screen for thyroid cancer, according to a study published in JAMA-Oto.
Researchers trained the algorithm on images from 121 patients who underwent ultrasound-guided fine needle-biopsy with subsequent molecular testing. The group then tested the model on a set of unlabeled images to see how closely it could classify high and low genetic risk nodules, compared to molecular test results.
The results showed that the machine learning algorithm performed with 97 per cent specificity and 90 per cent predictive positive value, meaning that 97 per cent of patients who have benign nodules will have their ultrasound read as “benign by the algorithm, and 90 per cent of malignant or “positive” nodules are classified as positive by the algorithm.
Mozilla partners with Element AI
On Monday, the tech giant said the strategic partnership between Mozilla and Element AI is focused on addressing "how new AI technologies and tools present challenges and opportunities for today's digital frontiers."
Element AI, an AI enterprise software provider that maintains existing partnerships with AWS, Microsoft, Nvidia, and Intel, will work with Mozilla to explore these aspects of ethical AI governance. The companies will also work on "data trusts," a new, proposed technological solution born from AI to measure and maintain data control, which may become key as AI works its way into data collection solutions.
Mozilla and Element AI hope to create data trust tools able to replace the current "broken consent-based system of data collection." The partnership will also involve funding and the support of AI-based legal and policy research.
Computer vision identifies ripe fruit and counterfeit drugs
Researchers are developing an application based on AI algorithms that works in regular smartphones and brings extremely accurate hyperspectral imaging within anyone’s reach. When would this avocado be suitably ripe for making guacamole? Is the drug I bought on my travels to far-off places the real thing or a fake?
Hyperspectral images are different from regular photographs because they reveal things unseen to the naked eye in the object photographed. The technique is not based on transillumination; rather, hyperspectral images interpret the wavelengths of light more accurately than regular photos.
However, the devices currently available are specialist equipment, with prices starting from several thousand euros. The less expensive technology developed by the University of Helsinki researchers could bring the solution to regular consumers.