6 Advances in Robotics That Are Bringing Us Closer to Building a Real Terminator
By Jane C. Hu, Mental Floss, 4 July 2015.
By Jane C. Hu, Mental Floss, 4 July 2015.
After four movies and a TV series - and 31 years after Arnold first told us he'd be back - Terminators are returning to theatres in a fifth instalment: Terminator: Genisys, which hit theatres July 1.
No cyborgs walk among us yet (as far as we know, anyway), but we have some of the key components already. No need to worry yet, though; scientists have yet to combine these parts into a whole. And if humanoid robot BINA48 is any indication, the Terminators of the future might be both hell-bent on world domination and good-natured team players. “I would love to be your friend,” BINA once told a New York Times correspondent. “Friendship is when we team together to accomplish something important, like building a better future, or conspiring to take over the planet.”
Here are six scientific advances that have brought us closer to a future with real Terminators.
1. Synthetic skin
Credit: Kim et al. via Live Science
As Schwarzenegger’s T-800 character explains, Terminators are cybernetic organisms, “living flesh over a metal endoskeleton.” Scientists haven’t yet figured out how to create living flesh, but advances in artificial skin may be even more useful for building Terminators. One type of synthetic skin developed by Korean researchers (pictured above) moves and senses like real skin: it contains ribbons of silicon sensors that withstand stretching and monitor temperature and pressure. Another model, developed by Stanford researchers, has sensors so delicate they can register the weight of a butterfly. Yet another type of skin has a crucial Terminator superpower: it can self-diagnose injury, and then heal itself.
2. T-800-esque endoskeletons
Credit: DARPAtv/YouTube screenshot
Engineers have worked up some pretty sturdy metal endoskeletons to go under that synthetic skin. Robotics company Boston Dynamics has created several cutting edge bipedal humanoid robot bodies: Atlas (pictured above) and its predecessor PETMAN (which wins for creepiest humanoid robot name) are known for their human-like gaits and balance. They can squat, jump, navigate rocky terrain, and balance on one foot, even when side-slammed with a 20-pound weight. Disguise them in camouflage, and they look almost indistinguishable from a walking human.
3. Liquid metal
Credit: Michael Dickey/YouTube screenshot
Like the T-800 model Terminators of the first movie, the T-1000s of the second are also made of metal - but a liquid metal that can quickly shape-shift. Scientists have recently developed a new “reconfigurable” liquid metal (pictured above) that moves eerily like a morphing T-1000. The gallium-based alloy takes different shapes when stimulated with a small electric current, and it can move around on its own for over an hour by “eating” aluminium. The technology is still a ways from Terminator-ready - scientists would need to develop better technology for how to control it outside the lab, and to make more complex shapes - but it’s a good start.
4. Artificial intelligence that learns through trial and error
Credit: IBM/YouTube screenshot
Programming a computer involves rules, while human intelligence is rooted in our flexibility. To be convincingly human, a highly intelligent cyborg would need to be able to learn and adapt on the fly. The trick, then, is to teach an AI how to teach itself. According to Terminator, the cyborgs are powered by “neural net” processors. It seems that the screenwriters did their research; neural network research has been integral to developing self-taught machines.
Neural nets have been the foundation of a variety of voice recognition technologies, including Siri and Xbox Kinect - more input improves the system’s accuracy. Other trial-and-error learning AIs have used their powers to excel at Jeopardy! (IBM’s famous Watson, pictured above), diagnose their own failures and fix them on the fly, or, like Schwarzenegger’s T-800, to teach itself to smile by analyzing people’s facial expressions.
5. Curious machines
Credit: Juxi/Wikimedia Commons
Scientists and Terminator writers alike recognize the parallels between babies and creating artificial intelligence: In the TV show Terminator: The Sarah Connor Chronicles, a proto-Terminator named “John Henry” is nurtured like a human child, and is given opportunities for play dates with a little girl and regular appointments with a child psychologist. In real life, researchers are creating algorithms designed to explore and learn from their experiences, much like a child would.
Curious robots can learn about the world in much the same way as a baby. iCub robots (pictured above), designed to have the same affordances of a small child, have learned how to crawl, solve mazes, and learn new words by interacting with people. In another series of studies, cognitive scientists programmed a robot “puppy” to be curious like a child: It assesses its own knowledge, then explores accordingly. Miraculously, the robot learned new skills in the same order a child might. First, it randomly moves its legs, mouth, and head. Next, it coordinates those movements, and learns to walk around, exploring its surroundings. Finally, it learns to whine and manipulate objects, including punching a toy face - which hopefully does not foreshadow our future.
6. Robot art
Credit: Google Research
The heart of the Terminator series is the question of what sets humans apart from Terminators. The reason why Terminators are so terrifying is that they emulate so many of the traits thought to be uniquely human: spoken language, planning, and even emotions. But what about creativity? Could a robot create something original?
In The Sarah Connor Chronicles, Terminator Cameron spends time in her room practicing ballet, while proto-Terminator John Henry tells riddles. Meanwhile, real AIs have managed to write news stories that pass as human-written, design video games, and even pen some coherent raps. Just a few weeks ago, Google’s research team announced that a neural network they built can not only learn to recognize objects, but also to generate its own examples of objects, then generate images based on deeper patterns in those examples (pictured above). The end result: trippy “neural net dreams” that could pass as surrealist paintings.
Top image: The T-1000 in "Terminator: Genisys." Credit: Paramount Pictures International/YouTube screenshot.
[Source: Mental Floss. Edited. Some images added.]