Looks like they've taken the criticism from Google et al. to heart. It was hard to see how the huge, gas-powered robots louder than lawnmowers were going to work for anyone except maybe military applications. Put some state of the art Google AI into this thing and it's not far from a sellable product!
These press releases from BD are certainly impressive and are always great fun to watch.
But they're mostly silent on the matter of how much autonomy the robots are operating with. I never know how much of their behavior is human directed, or how those directions are conveyed to the robots.
They say that robots are being remote controlled. So, there is an operator telling it to duck, grab, and where to go, but the robot does the actual balancing on itself.
I am a little bit disappointed that they are not using deep learning for robotics. Instead, they are simulating the robot based off a model, using Control Theory. It might be easier to get results from Control Theory but it doesn't offer a path towards more complex behavior, like Deep Learning.
Modern control theory is rather close to machine learning. Adaptive model-based feedforward control is machine learning. The machine learning part builds a model of the dynamics of the system. Then that model is inverted (solved for control inputs) to make it a control system.
They're doing this right. They have a very good basic body control system. Now someone can build higher level strategies to get work done on top of that. That's how biological brains work, after all. Google/Alphabet could, for example, reuse much of their automatic driving software as high level control for this robot.
Google should have BD manufacture a few hundred of those machines, and try to get the cost down to $25K or less per unit for that production run.
Mammal brains have multiple functional units. The cerebellum does most of the motor control. The cortex does most of the planning and deciding. The cortex acts through the cerebellum, not by driving muscles directly. Most of Boston Dynamics' control systems are doing cerebellum-level functions. As with the cerebellum, this involves fast control via feedback loops.
I would have to practice with the controls for a really long time to be able to pick up a soda can that deftly, though.
While they are not using deep learning, I think they're doing good work building a platform that can later be controlled via neural net or whatever. The control-theory-based balance and locomotion means that the lowest-level aspects of operation will not demand resources from the neural net and provides the prototypers in the meantime with a way of experimenting to see what body designs might be most effective.
Deep learning for robotics is tremendously harder than you're making it out to be. It's not magic pixie dust you just sprinkle over something. The state of the art in physical robot deep learning is robots learning a single simple action like picking up something in front of them, and even this is quite difficult.
Google is working on the deep learning aspect of finding out what to do / action and how to do it.
The control theory remark above is true, however even with deep learning, you can start your estimate of the parameters with the current control state, which reduces training time significantly.
If it really is using all electric actuators that's pretty big. This would be one of the first of Boston Dynamics' all electric robots.
What exactly does using electric actuators over hydraulics buy us? Less noise, greater efficiency, and reliability. Reliability is very important for both house hold and industrial robots. We typically measure reliability in terms of Mean Time Between Failures, aka, how long it typically last before breaking. For industrial robots this is important as the higher the reliability is the more money it makes. Industrial robots tend to have MTBFs of 100,000 hours or about 10 years.
Reliability is also important for household robots too, a big expensive robot that breaks down all the time appeals to few people.
The previous Spot robot was also all-electric. Hydraulic power was needed for the bigger machines. (Also, Raibert liked designing hydraulic systems; he has a patent on the valve/actuator combo used in BigDog.) Cube/square law - mass increases with the cube of the size. This is why insects have tiny leg cross sections in comparison to their length.
Battery life will be a problem, but it's clearly agile enough to plug itself in for a recharge.
Reliability is not an issue for industrial hydraulics (they rarely break down) however they do require a lot of maintenance, are filled with hot (dirty) hydraulic fluid and are noisy (pumps). Electric solves all those issues.
These robots are amazing in their realism of behavior that I find myself referring to them as creatures rather than robots.
Watching them get kicked makes me feel bad for them, just like I would for an animal, and watching them fall on a banana makes me laugh, just like I would if an animal or person did (so long as they weren't seriously injured, unless I didn't like them :P)
Google just said, "Wait! Did we say we are focusing on a 'household robot' and are therefore going to sell Boston Dynamics? Hehe. Just kidding about selling BD!" The cleaning up, stair climbing and fall recovery are seriously impressive.
We could have right now a RoboDog, with some machine guns on it and let them patrol and secure a perimeter. Fire at anything it moves. I hope they solve the issue with the banana peel :), though.
So much for Asimov's laws, eh? Autonomous weapons have had some problems [1] however. You need to be sure that no friendly fire is possible, and that's a tall order. I'd prefer Robo-sheepdog, myself.
Now that's a nice piece of machinery. Much closer to a salable product than the big machines they did for DoD. This is more in line with Google's business model. Maybe this is BD's effort to stay under the Alphabet umbrella. It will be good if it works.
They really have leg control and balance software figured out now. That machine is more agile than any of the previous BD machines.
Its really amazing to see how functional this robot is. I cant wait for one to take over for some of the more monotonous tasks around the house...
That said, the way it moves around and can keeps its head steady is both really cool, but somewhat terrifying... It seems to conjure up scenes from movies where the antagonist robots are scanning a target before deciding to kill or not.
They have an animated gif of playing tug of war with it! They are definitely trying to fix that whole "robot abuse" image that so many people got upset about last time!
Well if it gets out of control, it will probably fall over and not be able get up. If it malfunctions, probably the worst thing that could happen would be the battery pack catching fire, because of consumer safety laws there will be safeties in place to prevent this.
Hacking is an interesting one, only because current laws assume the perpetrator must be in proximity to the crime.
However, this robot is almost certainly not autonomous with the operator hidden behind the camera. Because it's not autonomous, it's going to be difficult to teleoperate over the internet. Telerobots are slow and not very capable[0]. With lag, it's going to be difficult to stab someone with a knife much less pick up a knife.
However, even slow telerobots present interesting legal issues. One might hack a telerobot in the home of some people who are on vacation and use it to unlock the doors so some guy can go in and steal stuff.
One of the best XKCDs ever is about this what-if scenario. It addresses full robot sentience -- if they suddenly became aware and tried to take over the world Terminator style. An "out of control / hack / malfunction" would be even more laughable:
How about in a case of a robot raping someone? Should I just let it rape me or someone near me until the batteries die?
Sorry, I was lucky enough to be born in a country where I'm allowed to defend myself. And if a robot or a person starts raping me, I'm not going to let them finish until they wear out.
Robots are currently bumbling electric imbeciles and do not perform well in such adversarial scenarios. They are also more fragile than they seem. One could permanently modify the robot's kinematics with the swift impact of an office chair or percussively decommission it's lidar sensor with a baseball bat.
/cdn0.vox-cdn.com/uploads/chorus_asset/file/6695725/giphy-_6_.0.gif){kind=link}
Looks like they've taken the criticism from Google et al. to heart. It was hard to see how the huge, gas-powered robots louder than lawnmowers were going to work for anyone except maybe military applications. Put some state of the art Google AI into this thing and it's not far from a sellable product!