The web giant is making a major push into artificial intelligence and robotics. What does this have to do with the search-engine business? And are robots on the battlefield the next step after driverless cars?
Google, it seems, wants to make everything in the world intelligent – and connected. Why else would the web giant spend billions of dollars to buy companies on the cutting edge of two fields poised to have an enormous impact on the future: artificial intelligence and robotics?
Users of all ages appreciate Google’s search engine; someday, probably sooner than most of us realise, they will be sitting in Google’s driverless cars and using Google’s intelligent vacuum cleaners, microwave ovens and washing machines.
Computer systems are already smart enough to interpret human speech or beat a Grand Master at chess. But until recently they have relied on pre-programmed rules to guide them through these tasks. More complex situations that involve millions of unexpected events – driving around a city, for example – simply could not be planned in advance. The solution to these more complex tasks is to train computer systems to learn for themselves, usually by analysing vast amounts of data.
One of Google’s most recent acquisitions could be crucial for achieving this goal. British company DeepMind develops algorithms for “machine learning”, a more advanced form of artificial intelligence. Some machine-learning strategies even mimic the brain by simulating its neural networks.
In 2011 Google used this approach to create a system that independently learned to recognise pictures of cats on the Internet. (Identifying cats may seem frivolous, but it is a good test of artificial intelligence because the system taught itself to recognise that their feline features identified them as a specific creature.)
Not everyone is convinced that Google just wants to make our lives better. Instead, many experts see a strategy aimed at protecting the company’s lucrative search-engine business. “That’s what pays Google’s wages,” says Mark Bishop, a cognitive-computing researcher at Goldsmiths, University of London. Google, he explains, is still dependent on the search algorithm that it developed in the mid-1990s, which, even though it has been improved, struggles to understand natural language. An example is a phrase like, “Time flies like an arrow; fruit flies like a banana,” which a human can parse easily by relying on context and experience.
What Google needs, according to Bishop, is a more powerful search tool that can “get at what people actually mean, rather than what they’ve written”. This in turn would enable the company to draw more nuanced conclusions about the interests and habits of its users – and target advertising accordingly.
The most publicised example of Google’s machine learning involves the driverless cars that it has been developing since 2009. The company uses a laser-ranging system to build a detailed map of the vehicle’s environment. Four U.S. states have authorised the cars to be tested on public roads, and the company is busily testing the system. Driverless cars herald a future with fewer accidents and less time wasted in traffic jams – but also more information for Google’s database on our personal lives.
Google’s acquisition of home-automation company Nest Labs offers another foothold in the “Internet of Things”. Nest Labs makes smart thermostats that detect when people are at home to control temperature, humidity and light levels, as well as smoke alarms that offer verbal advice on how to deal with any perceived threat. Both can be controlled remotely by Smartphone apps.
The next step will be to meet people’s needs automatically – for example, by ordering fresh food when a fridge is getting empty. “They’re not interested in your temperature, they’re interested in your data,” says computer engineer Pedro José Marrón of the University of Duisburg-Essen, Germany.
More worrisome are Google’s forays into the military field. Boston Dynamics, acquired in 2013, has developed BigDog, a four-legged beast of burden for the battlefield. It has also built Cheetah, which can sprint at 45 km/h, and a 2-m-tall humanoid search-and-rescue robot called Atlas. Recovering injured soldiers from the battlefield is one thing – but what if those robots were armed?
Thanks in part to lobbying by the International Committee for Robot Arms Control, with which Bishop is involved, the European Parliament has voted overwhelmingly to “ban the development, production and use of fully autonomous weapons which enable strikes to be carried out without human intervention.”
Google’s mission statement is “to organize the world’s information and make it universally accessible and useful.” As the company reaches into every aspect of people’s lives, many hope that it also remembers its informal corporate motto: Don’t be evil.
By Mark Peplow