Austria’s AI brain has plan to take down Waymo, Tesla
Think of the European automobile industry and countries like Germany come to mind, with its luxury cars whizzing down the speed-limit-less autobahn. Or France’s quirky Citroens and space-age Renaults. Or even the sensible Swedes and their sturdy Volvos.
Austria? Not so much. But the Alpine state is at the forefront of some cutting-edge research that stands to reshape the future of transportation, in particular driverless-car technology. It’s here that researchers from Alphabet Inc. to OpenAI, the artificial-intelligence group co-founded by Elon Musk, come calling for advice from Sepp Hochreiter, the head of the Institute for Machine Learning at Johannes Kepler University.
Hochreiter’s academic obsession since his university days three decades ago: artificial intelligence, in particular the development of long short-term memory (LSTM), a once obscure form of machine-learning software capable of processing sequential tasks. As a type of artificial intelligence loosely based on how parts of the human brain work, LSTM has made possible consumer innovations such as Amazon’s Alexa and other voice assistants. His research made Hochreiter a star in tech circles, though he’s politely resisted the Silicon Valley sirens.
The combination of brilliance and humility, captured in the 52-year-old’s slow-yet-steady ascent to the top of AI, could be Europe’s best chance of catching up to Silicon Valley.
“Tesla’s data is worth a lot and the company is very well placed,” Hochreiter said. “But there are different approaches to what Elon Musk and Tesla are doing. What we’re doing is better.”
With more than 90 percent of all Earth’s machine-readable information generated in recent years, Hochreiter calls data “the new oil” that drives the technology behind autonomous vehicles.
Musk’s company leads the field with an advantage the brash U.S. executive said in February “is very difficult to overcome.” More than a quarter million Tesla drivers are already behind the wheel, vacuuming up quadrillions of bytes of roadway data and transmitting it back daily to Silicon Valley.
Researchers like Hochreiter view cars as mini R&D centers that collect data needed to train autonomous-driving algorithms. And getting a critical mass of autonomous vehicles safely onto roadways is one of the “grand transformations” needed to cut greenhouse gas emissions and stave off catastrophic climate change, according to the International Institute for Applied Systems Analysis outside of Vienna.
Two wildcards are looming that could help change the race, according to BloombergNEF analyst Ali Izadi-Najafabadi. One is Europe’s impending completion of its Galileo satellite network, the 10 billion euro project that will improve the bloc’s access to high-precision positioning data.
The other is the onset of 5G telecommunications networks, which will allow autonomous cars to share data with each other, effectively turning them into data servers on wheels. Austria switched on Europe’s first commercial 5G network, with an emphasis on rural roadways where driverless cars can be tested.
Hochreiter certainly sees a home advantage. He’ll take the petabytes of driving information generated every month and combine it with other data sets — ranging from climate and pollution to social media and satellite feeds — that will allow them to construct new simulations of startling complexity, according to the scientist, who said Europe’s stricter regulations force him to take a different approach to teaching autonomous vehicles how to drive.
“Silicon Valley can be much more aggressive,” Hochreiter said. “They put cars on the road even when the system is not completely ready. Here the regulations are much stricter.”