How a brilliant teen uses high-tech AI to help others
I’m passionate about AI and deep learning. And I’m fascinated with combining AI and deep learning frameworks. AI isn’t just artificial intelligence anymore. It’s intelligence, augmented. Intelligence amplified. It’s taking what only humans can do and doing it even better. And when we combine AI with deep learning and neural networks, we humans can solve problems and do things that we’ve never been able to do before.
Why deep learning?
Let’s say I want to build a self-driving car. I need to tell the self-driving car to detect cars around it. But how will it know which objects are cars, which are light poles and so on? To create a library of images, I can easily put cameras in people’s cars and ask them to give me their camera information. But now, who is going to sit there and label every single frame individually, find every single car, draw boxes around them, then train the machine learning model? It won’t work. That’s simply impractical. And that’s where deep learning comes in.
Deep learning uses algorithms that permit software to train itself by exposing multilayered neural networks to vast amounts of data, for example, a huge library of images of cars, light poles and so on. Deep learning models can be trained to classify images and detect objects and to use auto-labeling to quickly label enormous numbers of images. The more the models learn, the deeper the neural network becomes, making more and more connections between images and what they represent, similar to how the human brain learns.
IBM PowerAI Vision helps SMEs create deep learning models
IBM® PowerAI Vision can help solve some of the key challenges with deep learning. It allows domain or subject matter experts (SMEs) to create their own deep learning models – without the need for coding or deep learning expertise. Users don’t need to know how the underlying GPUs work or how its multiprocessing works. They can focus on their deep learning models and pre-processing the data to run it through the system.
PowerAI software is powerful because it’s a toolkit that takes open source deep learning frameworks and packages them together on a Power system. The IBM Power Systems platform provides scaling efficiency and performance.
Deep Learning and healthcare
I believe the field where deep learning can have the greatest impact is the field of healthcare. I’m using Power AI and deep learning in a couple of different healthcare projects. In one, we’re building a neural network to help a young non-verbal woman communicate with the outside world. We’re developing deep-learning algorithms to help interpret and classify electroencephalogram (EEG) data gathered from the woman. We’ll use the results to help her mother and caregivers understand what she is thinking and what she wants instead of using their best guesses as they do now.
We’re also using deep learning and neural networks in a suicide-prevention and mental health chatbot application. We want to build AI-enabled bots that can supplement crisis lines and allow them to field more calls. A deep-learning based bot trained on Power AI can follow a free-form conversation and “remember” things a person said many sentences ago. We’re training the neural network in mental health so that the chatbot can help de-escalate crisis situations, prioritize calls and find a human to help as soon as possible. We’re making sure that even if there’s no human available at a particular time, there’s always someone that you can talk to. But we’re not replacing the human experts.
People wonder whether AI is meant to replace humans. This isn’t true at all. AI is meant to amplify our human skills, augment our intelligence and allow humans to do what we’ve already been doing, but do it in a much more efficient and accurate way.
Watch Tanmay Bakshi speak about using IBM PowerAI and deep learning to push the boundaries of AI: