Asimo is 12 years old, 1.30 meters tall, 48 kilograms, is Japanese and dresses like a little astronaut. Recognizes voices. Pour the juice from a bottle into a glass and serves a human. It is a formidable robot. But when his country faced the disaster of Fukushima, who stepped in to assess the damage of the largest nuclear accident since Chernobyl was not Asimo. It was a US military robots team much less friendly.
The ground came the miniature tank developed by iRobot for use in combat in Afghanistan. One of them - the PackBot, with the size of a laptop, mats for rough terrain and a mechanical arm with camera - filmed the interior of the plant and radioactive mapped points. A larger model removed debris up to 70 kilos. But the RQ 16a T-Hawk - an unmanned mini-helicopter - collected air plant images.
The moral of the story is that while the Japanese industry has spent decades to create humanoid robots able to do what we do, the Pentagon war machine invested in machines that do what we can not. The heart of this new chapter of robotics is the Research Agency Defense Advanced Projects, the Pentagon. In 1968, she had created the embryo of the internet. In the 80s, he invented the invisible fighters. And with the outbreak of the war on terrorism in 2001, he began to invest in robots. Why robots?
The reason is called asymmetric warfare. The war on terrorism, the fight does not happen between the armed forces of rival countries, but between a national army and inflitrados enemies among civilians. Then, supersonic fighters and atomic bombs developed in the Cold War are no longer important. Much more useful is a discreet eavesdropper robot that gathers information about the next place to attack.
The first robots of asymmetric warfare were the "drones" - unmanned aerial vehicles, children of model airplanes. Most of them used to field reconnaissance, surveillance and intelligence gathering. But there are also drones for air strikes, as the Avenger, invisible to radar and able to carry 2.7 tons of bombs. In eight years, drones were used in 302 attacks, leaving between 1,845 and 2,836 dead - 17% of them civilians, according to the NGO New America Foundation.
And the research continues. To achieve brilliant minds to design machines of war, Darpa launches calls the scientific community. In 2005, he offered $ 2 million for those who build a car that walked alone for 200 km on the mountain - 5 completed the route. In 2007, the challenge was that these vehicles will also follow the traffic signs. And in October will be the turn of the contest for robots with arms, legs, torso and head, able to without human control, handle common tools to cars. This to replace people in hazardous and degraded environments.
What does this have to do with our peaceful civilian life? Fukushima is only the first example. With a much lower cost than that of a helicopter, small drones can monitor traffic, cars chasing fugitives, scour forests and mountains in search and rescue operations and control borders. In the US, a survey of 1708 people found that 4 in 5 people support this kind of action. And it will come soon. The US Congress ordered the Federal Aviation Administration to create a plan for drones circulated by the civil airspace in 2015.
And intelligence?
As impressive as its capabilities, these robots are dumber than an ant. The drone pilot needs - even if it is on the ground. driverless cars follow pre-defined instructions and guided by GPS, never on its own. And cute robots like Asimo limited to follow a script. Already an ant looking for food alone and return home straight.
And the Deep Blue, IBM, which in 1997 defeated the greatest chess player in the world? Forget. "Who gave interviews to the press was defeated, since the computer could not speak," recalls Michio Kaku. Watson already knows. This robot, also from IBM, won in 2011 the best human contestants in the game tips and answers "Jeopardy!". He hears the tips in oral language, makes associations from millions of texts his 15 thousand gigabytes memory and responds to oral language. But can not perform the most basic task for a human: learning to interact with the world without a script.
This is so hard because computers process information in a completely different way to humans. On a computer, you provide a given, the processor transforms according to a pre-established rule, and the result is a new one. The rule is given up and down in series. In the mind the opposite happens. It is formed by neural networks operate not in series but in parallel. Hard to understand? Think then a baby. For him respond to a situation, various neural pathways are activated at the same time. As the practice, the neural pathway that lead to a better result will become stronger and stronger, and the other is will disable. That's how it is learning - a building rule from the bottom up, from the senses, experience and mistakes.
Theoretically a supercomputer would be possible to simulate the human brain, mimicking 100 billion neurons each connected to thousands of other neurons performing synapses at a rate of less than one millisecond. For this, one would first be necessary to 36.8 quadrillion computer calculations per second (petaflops). That we are not far - the most powerful supercomputer in the world, the IBM Sequoia, reaches 16.32 petaflops. Then it would be necessary to reverse-engineer the brain. Well, according to the futurologist Ray Kurzweil, this should happen in 20 years. Who lives will see. Or not.