One billion disabled people in the world
Around the world, many people have a disability — illness or injury — that affects their body or mind functions. The WHO (The World Health Organisation) estimates that 15% of the world’s population lives with a disability (one billion people). One-fifth of these, between 110 and 190 million people, have a severe disability. In 2017, in France: 1.5 million people were visually impaired and 850,000 suffered from reduced mobility.
World Report on Disability
About 15% of the world's population lives with some form of disability, of whom 2-4% experience significant…
The socio-economic situation of people with disabilities is marked by inequalities compared to the rest of the population. The challenge is double: develop accessibility but also increase the access to education, culture, work and training. From the development of the autonomous car to autonomous domestic assistants such as Amazon Alexa or Google Home, Artificial Intelligence is already helping to meet this challenge.
AI used to fight against all handicaps
In 2018, Microsoft has allocated a total budget of $25 million to fund projects focused on accessibility, based on artificial intelligence, over a five-year period.
Each selected company will receive a grant that includes free use of Microsoft’s Azure artificial intelligence platform. The diversity of the projects selected by Microsoft demonstrates the potential of Artificial Intelligence in the fight against disabilities.
Berkeley University’s Image and Video Processing Lab is developing an application that uses sensors and connected eyewear cameras to provide audio descriptions of the urban environment to visually impaired users. The research project combines Augmented Reality and Artificial Intelligence (Text recognition).
Augmented Reality for Visually Impaired People — D. R. Fox Design
Augmented Reality for Visually Impaired People (AR for VIPs) For the capstone research project of my Berkeley master’s…
The SuperVision Search application of Harvard Medical School University Hospital guides visually impaired users to the exact locations of desired items as shown in the following video.
Epilepsy Seizure Prevention
Scientists at the University of Sydney are developing a real-time brain signal processing system that can predict when the next epileptic seizure will hit a patient : 30 minutes maximum before the seizure with more than 80% accuracy.
Neuromuscular disabilities ALS (Charcot’s disease) and MS (Multiple Sclerosis)
The portable neuromuscular sensing device from Boston-based Pison enables users with neuromuscular disabilities such as ALS to communicate more easily through web micro-gesture-based control and platforms.
Disabled people on the fringe of AI applications?
AI can really change the daily life of people with disabilities. However, applications already marketed do not always work properly and can isolate them more.
Artificial Intelligence researchers and manufacturers should involve people with disabilities in the design of hardware and software that embed AI algorithms.
“I like to listen to music, especially old hits from the 60s. I ask my vocal assistant to play a particular song. He doesn’t recognize the title (or doesn’t like my voice) and plays a completely different song. I try to stop him, but he can’t recognize my voice from the music. I have to wait until the song ends and ask him to play the song I originally wanted to hear” — From Peter’s diary “Artificial intelligence and disability: too much promise, yet too little substance?
Need for more ethical data sets
Even more crucial than the design, the datasets that drive the algorithms must integrate data that take into account the disability.
Let’s take a concrete example: algorithms for computer vision systems are trained with images that take into account the attributes of faces, body posture or objects. However, disability can have an impact on a person’s body and physical appearance: on facial features, facial expressions, body, size or proportions, the presence of assistive equipment, movement properties.
All these elements must be taken into account when designing datasets used to train computer vision models. Built with ethics, these models and applications will also be used by people with disabilities.