From the human brain to machines: research on AI
Grady Jensen is both a neuroscientist and a computer scientist: the perfect combination for his current job as AI researcher at Volkswagen Data:Lab in Munich.
Artificial intelligence (AI) will play a decisive role in shaping the 21st century, whether we are talking about autonomous driving or robotics. That’s why the Volkswagen Group is hard at work in this field. The Volkswagen Data:Lab in Munich, the Group’s competence centre for AI, engages over 80 skilled international specialists. Grady Jensen is both a neuroscientist and a computer scientist: this combination makes him perfect for his job as AI researcher.
Understanding the human brain
He leads a team of eleven people, plus some PhD students. The results of their work is public and they’ve made their code open source so that anyone can work on it. The human brain holds the key to understanding and ultimately creating an artificial intelligence that can first learn and then act independently. “The brain is absolutely fascinating”, Jensen explains. “It can combine information and put it into a form that can itself think about the information and make predictions based on it – and all of this information comes from the body’s different senses”. Understanding how that works and being able to put it in mathematical terms is the most important basis for the development of powerful artificial intelligence.
Dynamic data analysis
Throughout the Volkswagen Group, more than 100 AI applications have already been implemented in vehicles, customer products and company processes. The main projects include intelligent robots able to work alongside people, and “bots” - self-learning systems that use intelligent data analysis to gain experience in order to perform repetitive tasks - as well as autonomous driving, of course. In what ways can programmers of self-learning systems learn from the human brain? One example: you’re sitting in a parked car. The cars to the left and right are moving, and for one moment you think you are also moving, even though you’re standing still; but after a short while you realise that you actually are not.
Establishing priorities
In the same way, artificial intelligence has to be able to manage the inputs and prioritise multiple simultaneous active data sources. For instance, for a robot that would be 360° cameras and pressure sensors. An autonomous vehicle, on the other hand, processes data from the radar system, GPS and other vehicles, and has to continuously evaluate which sensor input is relevant at that point in time. Although many may think that machines able to develop consciousness are just around the corner, we’re actually still very far away from that. “We still lack at the very least a solid mathematical basis”, says Jensen. “That’s fundamental, because artificial intelligence is connecting so many other fields of research by providing new algorithms and systems”.
Source: TOGETHER.net – Volkswagen AG