Google car crash casts doubt on robot powers
One of Google’s self-driving cars crashed into a bus in northern California. But the company insists that its investments in artificial intelligence will soon change the world.
Google’s driverless cars have travelled over 1.3 million miles since 2009. The technology is so advanced that in February this year, the computer system was recognised as a legal ‘driver’ by the US government. But just four days after the announcement, disaster struck: a self-driving car caused a crash for the first time.
It was driving around Mountain View, home of Google’s California headquarters. It went to make a right turn, but stopped when it detected some sandbags in the road. It changed lanes to avoid them — right into the side of a bus. Google’s car was moving at just 2mph, and no one was hurt. But it is the first time — despite 17 previous crashes — that the company has admitted their technology was at fault. ‘We clearly bear some responsibility.’
Learning to drive is a complex process. There are thousands of risks and signals, and the potentially erratic decisions of other drivers must be anticipated.
Google’s cars are learning all of this using a branch of artificial intelligence (AI) called ‘deep learning’. Traditional AI involved writing code to deal with every possible outcome. But deep learning mimics the way that humans learn — through experience. Essentially, it involves computers crunching huge amounts of data to teach themselves about the world around them.
Deep learning is not just about self-driving cars. If computers can master almost any subject by themselves, there are implications for almost every area of society. Right now, it is already curing diseases, studying climate change, and recommending the best TV.
But some fear that such powerful technologies could pose a far more serious threat. ‘Success in creating AI would be the biggest event in human history,’ said Stephen Hawking in 2014. ‘Unfortunately, it might also be the last.’
The crash in California was a setback for Google, but it continues to insist that its cars will be ready by 2020. And it is not alone; Elon Musk, founder of the competing driverless car company, Tesla, predicts that his vehicles will be ready in just two years. As AI gathers pace, who knows what the world will look like by then?
Calm down, say sceptics. People have been predicting that AI will be the best or worst thing to happen to humanity since the 1940s. But it never lives up to the hype. The phenomenon even has its own nickname: the ‘AI winter’. The revolution is a long way off — if it happens at all.
But this time it’s different, insist scientists. Just look at Go: people thought it would take another decade for AI to master the complex game, but Google’s AlphaGo is currently playing the grandmaster — and it might win. AI is far ahead of schedule, and it is only getting smarter.
- Would you trust a driverless car to get you somewhere safely?
- Will artificial intelligence be humanity’s undoing? Or its saviour?
- List five things that you think could be improved with artificial intelligence.
- There are many fascinating questions about the future of AI technologies. Re-write the final two paragraphs of the article above, posing two sides of a different argument.
Some People Say...
“We must not allow computers to become fully conscious.”
What do you think?
Q & A
- How will AI change my life?
- In a lot of ways, it already has — your Facebook newsfeed uses advanced AI tech to show you the posts it thinks you will be most interested in. If you have ever used Siri, you have communicated directly with an AI bot. This technology is only advancing — it could help solve a lot of problems. But then, it could also create them.
- Like robots taking over?
- Maybe. We are used to apocalyptic visions of AI gone wrong in science fiction. AI developers are aware of these fears, but they insist it is unlikely. More realistic problems may include mass job losses if medical or teaching professions become obsolete; privacy risks if an AI program accesses our health records or online habits; or the threat of cyberattacks from unfriendly forces.
- 17 previous crashes
- Until now, Google has said that these crashes were all due to ‘human error’ — usually because someone drove into the back of the driverless cars.
- A biotech company named Berg is using AI to comb through reams of patient data, hoping to find a cure for cancer. Another company, Enlitic, uses the technology to detect signs of cancer far earlier than humans could.
- Climate change
- Orbital Insight is a company which uses AI to locate water on the Earth’s surface, a system which could help to produce more accurate models of climate change.
- Netflix uses AI to learn what its users enjoy watching, and then ‘personally’ recommend the best TV shows and films.
- In January this year, Elon Musk said his electric car company had ‘all the pieces’ to create driverless cars. It just needed to refine them and put them together. In the future, he says, human-driven cars will be outlawed for being too dangerous.
- The traditional Chinese game is a bit like chess — but with more potential positions than there are atoms in the universe.