When I say “robot”, what do you picture? If it’s a humanoid collection of metal and wires, stomping around as it cleans your house or offers drinks to your guests, well yeah, that would probably be considered a robot. And if you’re imagining the opposite, with a Skynet-style intelligence destroying or enslaving humanity, well they’d be robots too.
But, like we’ve pointed out before, the word “robot” is about as vague a term as you can get, at least in engineering. They really are just machines designed to accomplish a task, which could describe pretty much anything; but outside of the dictionary we often take the word “robot” to mean ‘one of those humanoid computers cooking your dinner’. The thing is: there’s a whole lot more to robots than that. What we call a robot tends to have some combination of two elements: mechanical flexibility and artifical intelligence (or “AI”). That’s what it uses to accomplish its tasks. Depending on what they’re designed to do, a robot might need more flexibility or a weirdly specific type of AI. And some are more successful than others. With such loose terminology, picking out the most advanced robots in the world is tricky and probably also kind of impossible, since there’s no universally accepted test for how advanced a robot is. It’s subjective and your list might be different. But I’m going to give you my list anyway! Some of these are robots that already exist and are doing amazing things, and others are proofs of concept – hints of what the future will look like. The da Vinci surgical robot, for example, gets a lot of attention because it’s used all over the world. But it’s actually kind of weird to think of a da Vinci machine as a robot, because it doesn’t have much in the way of artificial intelligence. If you’ve seen that video of it stitching a grape back together, though, you know that it sure does have mechanical flexibility. With adjustable arms that can bend in ways that would break a human’s wrist, da Vinci is an incredibly flexible robot. It’s also a robot in the sense that it’s doing something, that previously could only be done by humans. It works by using two main parts: a station with a few arms that actually operate on the patient, and a place for a human surgeon to sit and control it remotely. You don’t just tell a da Vinci bot
“Okay, go remove the patient’s gallbladder now.” The surgeon controls every part of the actual surgery. and they’re equipped with special cameras that allow the doctor to see what’s going on at the surgical site, even through a tiny incision. It’s the kind of thing a human surgeon would never be able to accomplish, because there are limitations to what human hands can do. But the da Vinci’s capabilities have made some surgeries, like gallbladder removal, a lot less invasive than they used to be. Over the last decade, it’s changed the world of medicine and new versions are just getting better at what they do, giving surgeons greater control and flexibility. For example, newer models can be combined with fluorescence machines, where you dye the patient’s blood with a compound called indocyanine green. When it’s lit up by a laser, the dye emits green light, letting the surgeon more clearly see where the blood vessels are. Even though da Vinci has lots of programming to make sure it’s accurately translating the surgeon’s movements, it doesn’t need much AI, and in this case, you probably don’t want the robot to be making decisions on its own. For the next robot on our list, which is really just the latest in a long line of robots, AI is a lot more important. I’m talking about Curiosity, the rover that’s been exploring Mars since August 2012. Like da Vinci, rovers aren’t necessarily the first thing that might come to your mind when you think “robot”, but they’re jam-packed with AI, plus plenty of flexibility. Mars is far away. Depending on where Earth and Mars are in their orbits, it takes between 3 and 22 minutes for light to travel between the planets. So information isn’t being transmitted any faster than that. And it takes double the time to get a response, since you have to send a signal both there, and back. That’s why modern rovers aren’t just incredibly expensive remote control toys; it would be like trying to guide a remote controlled car, but having to wait half an hour every time you pressed a button. A huge waste of everybody’s time. Instead, every day the mission scientists send Curiosity a new sequence of tasks to do, and it goes ahead and does them, mostly on its own. One thing, that the rover has to be able to do really well on its own, is navigate. The team can tell it where to go, but it needs to able to avoid obstacles, like dangerous rocks or cliffs along the way. Curiosity uses no less than eight cameras to map out the shape of the terrain three meters in front of it to plot out a bunch of different possible paths, and then it chooses the safest one. One of NASA’s future rovers called “Mars 2020” is being designed on Curiosity’s specs, so it must be an effective system. But when it comes to navigating through hazardous terrain, some robots are designed to do a lot more than just roll over rocks. After the 2011 Fukushima disaster, DARPA started a challenge to motivate the best roboticists from all over the world to design robots that would be useful in natural disasters. In the DARPA Robotics Challenge teams compete to have the robots quickly complete a series of tasks, one of which is a surprise. They have to do things like climb over piles of debris, open doors, and drive a car, which, let’s face it, some humans don’t do all that well either. Mechanical flexibility is important, obviously, but they need some AI too, because the contest organisers figure that in a real disaster situation, communications won’t be too reliable. The tasks are so hard, that some robots don’t even complete the course at all. Some fall down and can’t get back up, and others just take longer than the time limit, which for this year’s challenge was an hour total for eight tasks. But the robot that took home the two million dollar prize did complete the course, in under forty-five minutes. The robot, called DRC-HUBO, was designed by a team from the Korean Advanced Institute of Science and Technology. Their strategy was to have the robot transform. It can walk on two legs, if you want it to, but it can also scoot along on wheels embedded in its knees. This way, when it was working on things that might have knocked it off-balance, like opening a door or drilling into a wall, it was a lot more stable. It also came equipped with plates on its legs, kinda like a bulldozer’s, so when it had to get past a pile of debris, it could just plow through it. Getting up a flight of stairs is also usually tough for a robot, because it needs to be able to see the stairs, but its knees get in the way.
That’s why HUBO walked up backward, spinning its torso all the way around, so that its cameras had a clear view of where it was going. AI was important, but it was creative solutions and the flexibility department that won this robot the competition. Self-driving cars, though, those are all about AI. Companies are developing fleets of them and the mechanical part is mostly figured out. They’re basically just regular cars, except with spinning sensors on top, that make them look kinda like submarines with whirling periscopes. The challenge is getting them to navigate highways and crowded city streets without banging into things – or people. And for that, you need lots of programming. The cars know all the rules of the road and are programmed to apply them to different situations, while avoiding hazards, like pedestrians, bikers and other cars. Google’s fleet has already logged over a million kilometers on the road, and it turns out they’re much safer drivers than humans, with only fourteen minor accidents since the project started six years ago. And none of them were the cars’ fault. It helps, of course, that these robot cars have split-second reaction times, never get tired or distracted by text messages, and they can share all those kilometers of driving experience. And you might be able to buy one as soon as 2017. Self-driving cars use a type of AI that analyses a situation and then decides what to do based on an incredibly complicated set of rules in its programming. Each car does that all on its own. They use a combination of cameras, radar and lasers to keep track of objects on the road, plotting the trajectories of other cars, bikes and pedestrians. That’s a lot more sophisticated than, say, a Roomba. But then there are robot swarms, which decide what to do together, using something called “Collective AI”. And in 2014, researchers at Harvard created the first real working versions, 1024 of them. They’re a little more futuristic than the other robots we’ve talked about, because robot swarms are still very much a new thing; they’re a totally different way of approaching AI and robotics. And now we know that it works. In Collective AI, the programming is deceptively simple. Like some colonies of biological organisms, the robots are coded so that they follow a series of simple rules. This particular swarm was designed to organise itself into shapes. So the rules are basically
“Figure out where you are, and if you’re on the edge of the swam, “move along until you find a spot where you’re allowed to be, that helps make that shape.” That makes each individual bot a lot simpler to program, and all of them can work together, without having to actively communicate. Eventually, this kind of technology could be used in fields like construction or medicine. If someone had an infection, for instance, you could just send a fleet of tiny robots into their blood stream and have them seek out and destroy the virus or bacteria. A swarm of tiny, simple robots, or a set of surgical arms might not quite fit in with the Terminator, but when it comes to the most advanced robots in the world, they’re on my list. Which robots would you have included in your list? Let us know in the comments below. Thanks for watching this episode of SciShow, which was brought to you by our Patrons on Patreon. If you want to help us keep making videos like this, check out patreon.com/scishow, and don’t forget to go to youtube.com/scishow and subscribe.