Microsoft has developed a new smartphone app that interprets eye signals and translates them into letters, allowing people with motor neurone disease to communicate with others from a phone.

The GazeSpeak app combines a smartphone’s camera with artificial intelligence to recognize eye movements in real time and convert them into letters, words and sentences.

For people suffering from ALS（渐冻症）, also known as motor neurone disease, eye movement can be the only way they are able to communicate.

“Current eye-tracking input systems for people with ALS or other motor impairments are expensive, not robust under sunlight, and inconvenient to operate and carry,” said Xiaoyi Zhang,a researcher at Microsoft who developed the technology.

“To overcome the drawbacks, we created GazeSpeak, which is designed to be low-cost, robust, portable and easy to learn.”

The app is used by the listener by pointing his smartphone at the speaker. A chart that can be stuck to the back of the smartphone is then used by the speaker to determine which eye movements to make in order to communicate.

The sticker shows four grids of letters, which each corresponds to a different eye movement. By looking up, down, left or right, the speaker selects which grids the letters they want belong to. The artificial intelligence system is then able to predict the word or sentence they are trying to say.

Zhang’s research, Smartphone-Based Gaze Gesture Communication for People with Motor Disabilities, is set to be presented at the Conference on Human Factors in Computing Systems in May.

Some roboticists believe robots can actually solve some of our biggest social problems, including loneliness and dementia (痴呆).

In 2019, Japanese robotics company GrooveX introduced LOVOT, a robot that weighs as much as a baby and looks like a mix between a penguin and a bear. In fact, LOVOT is just one of the latest robots meant to help people who are suffering from issues such as loneliness and dementia.

LOVOT has a system of sensors that allows it to move freely around a room and respond to things. For example, they will get its belly rubbed or fall asleep when it’s held. Also, when LOVOT needs a hug, it will find its owner in the house and wait until it is picked up. All over the world, robots like LOVOT are being used for “robot-assisted therapy (治疗)”. For example, when a robot baby seal named PARO was given to an elderly Australian patient with dementia, the patient spoke for the first time since arriving at the nursing facility. In New Zealand, when dementia patients were given the chance to play with PARO or a real dog, they chose to play with PARO.

The invention of robots like LOVOT and PARO is no accident. Our society is quickly changing, and some countries are becoming “aging societies”. Sometimes, these older populations don’t have enough love or companionship in their life, and their age or health might mean they can’t care for a pet. That’s why LOVOT and PARO could be the perfect companions for both young and old folks, always ready to give love without ever asking or anything in return.

Machine learning is a way of teaching computers to do tasks. In the past, to do anything, computers depended on detailed, step-by-step instructions written by a person. But with machine learning, computers pick up new skills on their own, using examples or experience. This makes machine learning a powerful form of artificial intelligence, or AI. AI is the ability of a machine to think or learn like a person. AI systems power many kinds of machines — from smartphones to robots.

There are a few types of machine learning. The most common is supervised (监督的) learning. In this method, a computer is fed lots of sorted training data (数据). For example, it may look at many photos of dogs. Each photo might be sorted by the dog’s kind. By studying those photos, the computer learns what each kind of dog looks like. Then, the computer can pick out dog kinds in new, unsorted photos. The more data the computer studies, the better it gets at its task.

Another important kind of machine learning is reinforcement (强化) learning. Using this type of learning is sort of like training your dog to do a trick (把戏). The computer learns to achieve some goal by interacting with its environment. When the computer makes a choice that brings it closer to its goal, it gets a virtual reward. Through trial and or, the computer learns to make better choices. This type of machine learning has helped computers master difficult games like Go and StarCraft II. It has also been used to teach self-driving cars how to get around.

Machine learning has helped create technology all around us. It powers language translators and search engines. When your streaming services recommend new shows you might like, that’s machine learning at work. Machine learning systems also organize posts in social media feeds. They even help judge diseases. In many ways, the problem-solving power of machine learning is reshaping the world.