Munich, Germany - Sometimes you have to go back to basics. And that’s what they’re doing at the Volkswagen Information Technology Centre: asking fundamental questions about intelligence.
Their research is focused on how people think, on what basis they take decisions, and how artificial intelligence can help them in the future. Because what makes us human is our ability to learn, to add two and two and get four - a number that wasn’t there before.
Volkswagen’s research team includes IT experts, robotics specialists, data scientists, programmers, physicists and mathematicians - that’s how complicated intelligence is. What you see in the lines of code graphs and 3-D diagrams on their computer monitors is machine language that could one day help accelerate corporate processes, optimise traffic flows and guide autonomous vehicles safely through the traffic.
We’re developing learning systems,” explained AI research head Dr Patrick van der Smagt. “We’re exploring and creating algorithms that can identify and predict patterns more reliably."
These are called deep neural networks, where the algorithm gains information in several steps and compares it with what has already been learnt, to construct a reliable rule - an ‘always’, to use the human term. A simple example is: If it rains, you always get wet. And there are algorithms already in existence that don’t have to be told that - they can figure it out for themselves.
A few doors away in the same building, theory is becoming practice. Dr Hakan Duman and his team are testing those AI systems to see where they can be used effectively, and what everyday processes and tasks are suited for the very straightforward way AI systems think. You wouldn’t ask an AI system to write a novel, he said, although once it had read enough of them it probably could, but you could ask it to edit a novel written by a human. It’s called ‘spell check’.
Duman's projects include AI systems for intelligent robots that, instead of performing set tasks in a given sequence, over and over again, can work alongside humans and learn from them. Another project deals with the next generation of traffic flow optimisation, using data as diverse as weather patterns, what day it is, and the dates of school terms. The team is also working on the possibility of an intelligent computer firewall that could protect sensitive data from malware that it’s never seen before.
Volkswagen chief information officer Dr Martin Hofmann explained that artificial intelligence was not an end in itself, that there were ethical aspects involved.
“Artificial intelligence should always help human beings in a meaningful way," he said, adding that this was why Volkswagen had adopted an open-source approach. Major sections of the software are made available to the public; in that way the development work carried out by the specialists is transparent and can be verified.
And if it doesn’t work as expected, they do just what you and I do in similar circumstances: press the ‘reboot’ button.