The recent release of Terminator: Dark Fate saw both Arnold Schwarzenegger and Linda Hamilton reprise their iconic roles and James Cameron's return as a writer and producer. While the sentient killing machines depicted in the Terminator franchise are comprised of CGI and Hollywood special effects, plenty of real world research is going into developing robots with similar capabilities, just without the murderous intent (we hope).
This research spans academia, militaries (though it can be difficult to suss out the actual breakthroughs from government propaganda), and private enterprise. Perhaps the most well known privately-owned robotics developer is Boston Dynamics, makers of the Atlas. You may remember this bipedal robot from September when it showed off its uncanny parkour abilities, which the robot can pull off 80 percent of the time. The Atlas is able to move so fluidly thanks to a novel optimization algorithm that breaks down complex movements into smaller reference motions for its arms, torso, and legs. The Atlas then utilizes a model predictive controller to chain each appendages movements into smoothly flowing movements. However, while Boston Dynamics' Big Dog was developed as a quadrupedal cargo carrier for military operations, the Atlas is strictly for use as an emergency first responder.
But for all of Atlas' fancy footwork, it doesn't look or work very much like the humans it aims to imitate. But, then again, neither did the T-800 from The Terminator and T2: Judgement Day -- at least once stripped of its biological covering. As you can see in the clip below, the T-800's muscles don't operate like a human's. Instead of bundles of contracting fibers, it utilized a complex series of delicate pneumatic compressors to manipulate its movements.
However, building bundles of synthetic muscles is exactly what a number of researchers are currently attempting. These fibers can be made from a variety of materials, from carbon fiber to nylon to exotic polyethylenes. When activated, these materials are capable of lifting up to 1,000 times their own mass (far more than we can) as well as retain a "memory" of their previous shape.
For example, a team of researchers from MIT developed a polymer that can expand 1,000 times its original length and pick up 650 times its own mass. It does this by bonding high-density polyethylene (the stuff used to make recyclable soda bottles) and a stretchy elastomer. This bonded pair naturally coils, like a bedspring. But when heat or cold is applied, the HDPE expands or contracts at five times the rate of the elastomer which lengthens or shortens the coil by as much as 50 percent of its original length.
Similarly, a team from Columbia Engineering recently developed a 3D printed synthetic muscle that not only expands and contracts but also bends, and even twists, on command -- while hauling 1000 times its own mass. This is a big deal because, like the T-800, today's robots are mostly driven by pneumatics, which severely restricts their applications and their overall size. This material, however, can be activated with just 8V of current.
"We've been making great strides toward making robots minds, but robot bodies are still primitive," lead scientist Hod Lipson said in a 2017 statement. "This is a big piece of the puzzle and, like biology, the new actuator can be shaped and reshaped a thousand ways. We've overcome one of the final barriers to making lifelike robots."
Electricity isn't the only potential source of power for these synthetic fibers. Human muscles run on glucose and oxygen, so why not a robot's as well? A research team from Linköping University, Sweden recently did just that and published their findings in the journal, Advanced Materials, this past June.
Their muscle consists of two electroactive polymer sheets sandwiching a non-conductive central membrane. When a positive current is applied to one side (causing it to contract) and a negative current is applied to the other (causing it to expand), the entire thing bends towards the positive charge. But rather than use an electrical current, the Linköping team integrated a naturally occurring enzyme capable of converting chemical energy into electrical energy.
"These enzymes convert glucose and oxygen, in the same way as in the body, to produce the electrons required to power motion in an artificial muscle made from an electroactive polymer. No source of voltage is required: it's enough simply to immerse the actuator into a solution of glucose in water", Edwin Jager, senior lecturer at Linköping University, said in a June statement.
Skin is another sticking point for the T-800 -- it can't travel back in time without an "Edgar Suit" after all -- but modern research is already hard at work on growing human skin in the lab. Don't worry, we're not bringing back Leatherface. It's actually to help eliminate the need for animal testing in the cosmetics industry.
In 2015, cosmetics giant L'Oreal teamed up with 3D printing startup Organovo to begin bioprinting human skin, in half-centimeter square patches. Similarly in 2016, the RIKEN Center for Developmental Biology paired with Tokyo University to grow a nearly complete epidermis -- down to the hair follicles -- that could be transplanted onto live subjects and actually work. The team took cells from the gums of mice and reset them to their stem cell-like iPS state before culturing and then implanting them on other mice, where they grew into integumentary tissue -- that's the layer of cells between the outer and inner skin layers where hairs are developed.
But the skin suit does not make the man -- sit down, Buffalo Bill -- at least when it comes to Terminators. It's their big beautiful AI brains. Obviously, we don't have anything as capable as what sits between the T-800's audio inputs, but that doesn't mean we're not trying. Many of the biggest names in Silicon Valley, including Apple, Huawei, Qualcomm and Alphabet, are racing to develop a new generation of processors specifically designed to handle machine learning tasks. Similar to ARM chips, which pair slower-performing but less energy-intensive cores with more powerful cores with a bigger current draw, the latest generation of "AI chips" integrate cores dedicated to machine learning functions. Image recognition and those sorts of applications -- looking at you Apple Face ID -- are instead routed to the GPU's neural engine.
The T-1000 (portrayed by Robert Patrick in T2 and Byung-hun Lee in Genisys) conveniently didn't require a flesh jacket to get back through time on account of its mimetic poly-alloy "liquid metal" construction. In the movies, these poly-alloys enabled the T-1000 to shrug off immense amounts of damage and change its shape at will. Real-life liquid metals like gallium offer some unique properties of their own like high electrical conductivity and deformability. But there are drawbacks. Most magnetic liquid metals suffer from a high surface tension, limiting their stretchiness to just the horizontal plane. Plus, they typically have to be submerged lest they become a sticky paste when exposed to atmosphere.
To get around these issues, a team of researchers submerged a droplet of gallium-indium-tin alloy in a hydrochloric acid bath. The gallium alloy reacted to the acid, forming a gallium oxide skin on the droplet, which drastically lowered its surface tension and allowed the droplet to be stretched both horizontally and vertically. The results of the team's experiments were published in the journal Applied Materials & Interfaces this past March. But don't worry about getting a finger needle through the eye anytime soon. This research is still in very early development, though it could one day find use in flexible electronics and soft robots.
We're also not likely to see robots melting through gates in the near future, but plenty of robots can already modify their shapes in response to environmental changes. NASA, for example, is working on the Shapeshifter ahead of a proposed expedition to Saturn's moon, Titan.
"We have very limited information about the composition of the surface. Rocky terrain, methane lakes, cryovolcanoes - we potentially have all of these, but we don't know for certain," JPL Principal Investigator Ali Agha said in a statement. "So we thought about how to create a system that is versatile and capable of traversing different types of terrain but also compact enough to launch on a rocket."
The team's answer is a gang of up to 12 small robots, dubbed "cobots," that can Voltron themselves into various configurations depending on the challenge at hand. Each would be capable of autonomous flight. Together they'd be able to daisy chain themselves across gaps or combine into a large wheel for faster overland travel. The team plans to submit their proposal in 2020 for consideration ahead of the next scheduled mission to Titan in 2026.
It may not be able to fully recombine on the other side of a security gate, this tendril-like robot developed by UCSB and Stanford researchers can easily make it between the bars. Taking inspiration from the movements of plant and fungal roots, the inflatable robot can extend up to 72 meters in length. Think of it as one of those balloons that clowns twist into animals, just 236 feet long. By incorporating specialized "control chambers" the robo-tube can also change direction, manipulate objects and even form its own tools, like hooks.
In the third Terminator, T3, Skynet has improved upon the T-1000's poly-alloy design -- this time using it as a protective coating over a super strong endoskeleton -- to create the T-X model. It doesn't just hunt humans, the T-X is a Terminator-killer to boot.
Unfortunately, plenty of research has already been sunk into developing autonomous war machines. In 2016 the US Navy and DARPA collaborated on the Sea Hunter, an autonomous anti-sub system, the US Army is currently accepting proposals for its Advanced Targeting and Lethality Automated System (ATLAS), an AI-powered system able to "acquire, identify, and engage targets at least 3X faster than the current manual process," per the solicitation notice. The Air Force is also exploring the idea of fully autonomous drones as part of its Skyborg project. And those are just a few of the programs we know about.
Whether these systems ever see the light of day -- at least publicly -- remains to be seen given the tremendous public outcry against autonomous weapons. Human Rights Watch is a founding member of the Campaign to Stop Killer Robots and calls for a "preemptive ban on the development, production, and use of fully autonomous weapons." In 2015, robotics researchers and tech luminaries like Steve Wozniak and Stephen Hawking penned an open letter arguing against their development.
"You can't have machines deciding whether humans live or die," Toby Walsh, a professor at the University of New South Wales, told the NYT in July. "It crosses new territory. Machines don't have our moral compass, our compassion and our emotions. Machines are not moral beings."
These pleas have not gone unnoticed. Earlier this month, the Pentagon released draft guidelines regarding AI development. The guidelines demand that AI systems be accountable, avoid bias and be "governable." That is, the systems use an inhibitor function to stop themselves before causing unnecessary harm or damage. Then again, on November 5th, the bipartisan National Security Commission on Artificial Intelligence called for the rapid development and deployment of autonomous weapon systems -- ethical concerns be damned.
"In light of the choices being made by our strategic competitors, the United States must also examine AI through a military lens, including concepts for AI-enabled autonomous operations," the commission's interim report reads.
What could possibly be more terrifying than an unstoppable killing machine? An unstoppable killing machine that can step out of its own skin to become a pair of unstoppable killing machines, that's what. And that's exactly what Sarah Connor has to defeat in Dark Fate. The Rev-9 Terminator builds off of the T-X's endoskeleton-wrapped-in-liquid-metal design except it can separate its halves and fight like those freaky blonde twins from Matrix Reloaded. Our current state of the art swarm technology can't coordinate at that level just yet, but it's getting close.
— James Kimbrough (@JamesKimbrough) February 6, 2017
Drone swarms can actually be quite useful by splitting sensory and processing functions across a group of robots. Lady Gaga would never have been able to pull off her 2017 Super Bowl Halftime Show were it not for a swarm of 300 Intel drones.
This technology has also caught the attention of the US military. The Army, for example, has developed the Perdix system, a hoard of more than 100 microdrones which are launched from a trio of F/A-18 Super Hornets and provide low-altitude surveillance for troops on the ground. The US Navy is developing a similar system, dubbed "swarmbots." These autonomous patrol boats coordinate with one another to investigate suspicious vessels that enter their domain (in this case Chesapeake Harbor) and then relay that information back to a human supervisor. But not all drone swarms wear capes. In 2018, a kidnapping ring leveraged a swarm of drones to buzz an FBI hostage team in the middle of their operation and keep tabs on the Feds' movements.
So even as the military and private enterprises continue to slog towards a future filled with autonomous weapons of war, we can take comfort in knowing that at the current rate of development, we likely won't face a Terminator uprising in our lifetimes. Then again, those maniacal mechanical bastards can time travel.