Leveraging Artificial Intelligence to Create Impressive Products for Disabled People

Technology is an excellent way to enhance the lives of people with disabilities. With the advent of artificial intelligence, several avenues of research have opened up that focus on enhancing the lives of people with impairment.

For instance, Facebook has designed an AI tool that can help the blind “see” again. This AI model explains the images on the Facebook feed of a blind person, so the person using the screen reader gets an idea of what is going on in the picture. This means people with visual impairment no longer have to hear a screen reader say “Photo” by “John Doe.” Google’s ‘Look to Speak’ app uses machine learning and computer vision to allow users to control their devices with their eyes

Similarly, OrCam, a Jerusalem-based company, has developed an AI-based called OrCam Read. This handheld device can read full pages or screens of text aloud from any printed or digital surface, including newspapers, books, product labels, and computers and smartphones. Through this device, OrCam aims to help people with reading challenges, such as dyslexia, mild to moderate vision loss, reading fatigue, as well as for those who read large volumes of text.

Even company giants like Microsoft have started a five-year program called ‘AI for Accessibility,’ with an investment of US$25 million, aiming to put AI in the hands of developers to make the world more accessible by providing AI solutions for the specially-abled. Artificial intelligence not only assists people with physical disabilities but is also helping people struggling with learning problems and mental health issues. E.g., Microsoft’s Windows Hello uses biometric login, i.e., fingerprint, face, or iris, which can work for people with physical disabilities or those with dyslexia who might struggle to remember passwords. AI chatbots like Woebot and Wysa are ensuring the availability of consultation for mental health woes, beyond the therapist hours 24/7.

Meanwhile, people suffering from epilepsy can have seizures from blinking lights and animations. This is why accessiBe, a web accessibility platform enables epileptic users to disable various types of animation, such as GIFs and videos so that they can browse the web without complications. Voiceitt is an app for people with speech impediments, including both those who need it temporarily after strokes and brain injuries, and those with more long-term conditions like cerebral palsy, Parkinson’s, and Down’s syndrome. The app uses machine learning to pick up speakers’ unique speech patterns, recognize any mispronunciations, and rectify them before creating an audio or text output. Livio AI, developed by Starkey, an AI medical device company, is a hearing aid that will enhance the hearing experience by quieting all the external noise from the environment and tracking health-related data to enable patients to seek help during emergencies.

Thanks to artificial intelligence, autonomous vehicles also promise to provide people with disabilities more mobility than ever before. Once the self-driving vehicles are fully integrated into society, they can be a resourceful asset for people with different disabilities, including motor impairment. These people would no longer be dependent on other people or public transport.

Further, most of the existing testing methods are highly ineffective at pinpointing learning disabilities like dyslexia or dyscalculia. Artificial Intelligence can help teachers and healthcare professionals diagnose early signs of such conditions and help the students accordingly. For instance, Australian startup Dystech has developed a screening app for early detection of such learning disorders.

Built on Amazon Web Services (AWS), Dystech employs artificial intelligence and machine learning to screen test if the user has dyslexia or dysgraphia. For the former, the app uses datasets of audio recording from both dyslexic and non-dyslexic adults and children to train the AI and relies on users reading aloud words that appear on the screen while being recorded using their smart device during assessment . And for dysgraphia it uses a photo of a handwritten text for screening. After subjected to a 10-minute screening test, app informs users about their likelihood of having dyslexia or dysgraphia.

The post HOW IS ARTIFICIAL INTELLIGENCE TRANSFORMING THE LIVES OF PEOPLE WITH DISABILITIES? appeared first on Artificial Intelligence.

Microsoft’s Azure Machine Learning team recently open-sourced their contribution to the ONNX Runtime library for improving the performance of the natural language processing (NLP) model BERT. With the optimizations, the model’s inference latency on the SQUAD benchmark sped up 17x.

Senior program manager Emma Ning gave an overview of the results in a blog post. In collaboration with engineers from Bing, the Azure researchers developed a condensed BERT model for understanding web-search queries. To improve the model’s response time, the team re-implemented the model in C++. Microsoft is now open-sourcing those optimizations by contributing them to ONNX Runtime, an open-source library for accelerating neural-network inference operations. According to Ning,

With ONNX Runtime, AI developers can now easily productionize large transformer models with high performance across both CPU and GPU hardware, using the same technology Microsoft uses to serve their customers.

BERT is a NLP model developed by Google AI, and Google announced last year that the model was being used by their search engine to help process about 1-in-10 search queries. BERT is useful for handling longer queries, or queries where short words (like “for” and “to”, which are often ignored in standard search engines) are especially relevant to the meaning of the query. Bing also began using deep-learning NLP models in their search engine last year. However, BERT is a complex model, which means that processing a search query through it (aka inference) is computationally expensive and relatively slow. Bing found that even a condensed three-layer version required twenty CPU cores to achieve 77ms latency, which is already close to the limit for users to notice a delay. To handle the volume of queries at Bing’s scale, using this model would require thousands of servers.

BERT inference does benefit from the parallelism of GPUs, and Bing found that the inference latency on Azure GPU VMs dropped to 20ms. For further improvements, the team partnered with NVIDIA to re-reimplement the model using TensorRT C++ APIs and CUDA libraries. This optimized model achieved 9ms latency. By using mixed precision and Tensor Cores, the latency improved to 6ms.

Deep-learning inference performance is a major concern, for web searches as well as mobile and edge devices, but re-implementing models by hand is not an attractive solution for most pracitioners seeking to improve performance. Inference acceleration tools, such as TensorFlow Lite and PyTorch Mobile, are now standard components of deep-learning frameworks. These tools improve performance by automatically re-writing the model code to take advantage of device-specific hardware acceleration and optimized libraries. This process is very similar to that used by an optimizing compiler for a high-level programming language, and similarly requires an abstract representation of the model being optimized. ONNX is an open standard for such a representation, and ONNX Runtime is an implementation of the standard.

Taking the lessons learned from re-implementing BERT, the Bing and Azure devs updated the ONNX Runtime code to automatically optimize any BERT model for inference on CPU as well as GPU. When used on the three-layer BERT model, CPU performance improved 17x and GPU performance improved 3x. Bing developers also found the ONNX Runtime was easier to use and reduced their time to optimize new models.

BERT model optimizations are available in the latest release of ONNX Runtime on GitHub.

The post Microsoft Open-Sources ONNX Acceleration for BERT AI Model appeared first on Artificial Intelligence.