Heather Knight, a roboticist and founder of Marilyn Monrobots, is trying to bridge the uncanny valley by adding humor to the robotic repertoire. Her robot, Data, can do imitations of Darth Vader, R2D2, and Buzz Lightyear. (Source: Freescale Semiconductor)
Definitely an interesting question, Alex, and one I'm sure you'll get plenty of feedback on. I agree with the sentiment expressed in your post that people are essentially human programming machines, thus more likely to mimic their own behaviors and patterns in the robots they design especially as those robots get more sophisticated.
Still, I think there's got to be a place for both kinds. For industrial applications deep in the factory or in out in space, for example, there's no reason why a robot should exhibit any ressemblance to humans. However, for applications where there is heavy human interaction, then it's probably quite comforting and even more productive to deal with a machine that has human-like traits as part of enhancing the collaboration experience.
Humans are genetically programmed to see faces...the kind of faces hiding in the bushes that are friendly, spying, about to attack, or have a mouthful of sharp fangs that wish to devour you. Since we see faces in just about everything anyway, why not design a face or humanoid style into the product on purpose. At least that will allow the designers to control the "happy" or "aggressive" perception that users get when they see their product. Rather than, for example, purchasing an automatic rocking chair that just happens to look like a hungry tarantula...
That's a great point, Bill. Although I "knew" it, I realize now that I've never really thought about that explicitly. But it's correct and I'd add to it that I've always thought everyone has a facial recognition program embedded in their brain as an instinctual function. I.e., it's like real-time software in that whenever one sees a new face (human or animal), this program performs the "friend or foe" analysis. That's what early man had to figure out for survival. I guess by extension we recognize robots as robots by perceiving the lack of movement/life in the faces of those humanoid machines. Conceivably, manufacturers could play with like/dislike by altering the characteristics (faces) of those robots.
I don't have a strong opinion on the aesthetic, psychological, or cultural implications of humanoid robots. But there are practical reasons for employing aspects of human form and function (as well as the form and function of other living things) in robot design -- and engineering design in general.
Over a timescale of billions of years, living things have evolved all kinds of interesting solutions to mechanical problems, such as locomotion, which I strongly doubt any engineer would be intelligent or creative enough to come up with on his or her own. Living things also give us examples of self-assembling, nanostructured, multifunctional materials, such as bone or spider silk, which rival any of our current industrial materials. That's not to mention sensing methods, self-organization, etc., all of which have been perfected by living organisms.
When I look back on my education, one thing which I really regret is the fact that I never took a single biology class after my freshman year of high school. I'm trying to remedy that now, with an MIT biology class which is freely available on OpenCourseWare.
I agree, Dave. Robots may mimic human actions or the actions of other living things. The most successful use of robots so far seem to be in the automated production of goods. These robots mimic hand and arm motions somewhat, but that's coincidental. Ultimately, robots will be successful as they reduce costs, improve performance, or perform in areas humans can't reach, such as in space or toxic environments.
Mention "robot" and many people think of something from "Star Wars" or "The Day the Earth Stood Still." Perhaps it's better to think about automation. When people wanted to create an automated piano they didn't create a mechanical "humanoid" to press the keys and replace the human player. Instead they used a roll of paper punched with holes to control the keys. Automation in Detroit took a different approach--human-like robotic arms that sprayed paint and welded metal. You design the automation to fit the task. If I want an aquatic robot that can swim quickly, it will look like a porpoise and not like a human or a movie-prop "robot."
Many people might feel comfortable around human-like robots, but they all look a bit creepy to me.
I agree Jon. Your porpoise analogy is a good example. Automation is a good name for it. A good portion of consumer goods are effectively made by robots, since the automated line is really a robot. The preference consumers have for the automated teller machines is proof that we don't need to dress up our automation to look like people.
Jon, I like your porpoise example. This question presupposes that humans are correctly constructed for our purpose, what ever that is. Obviously we are not or we would not have to use tools to perform virtually any task. If my robot is to lift and rotate an automobile engine what human do I use as a model?
Thanks, Tool_maker. Yes, we humans are rather frail. Backs, knees, and hips seem to require regular maintenance, and in the case of hips and knees, replacement. Robots go through shorter evolutionary changes than humans.
Good point, Tool-maker. As automation becomes more sophisticated, I think it replicates human physical action less and less. At a certain point, automation efficiencies requre thinking outside human movement. Continual improvement seems to take a non-human-movement path.
That's a good question, Ivan -- especially now that Google had commandeered the term Droid. What I found is that an android is an automated system that resembles humans or has human characteristics. A robot is simply an automated system. Though I also found definitions for robot that said robots have human characteristics.
The Japanese affinity for making robots look human is as humorous as those who feel Sheedzu's (spelling), a dog bred in Tibet quite capable of handling cold, need to be clothed. Isn't it funny that people want to make things that they love like people; I think it is a reflection upon the lack of such love between people. However, it does put a smile on my face to see such silly things, even if they're not intended to be silly.
I was also pleasantly surprised the other day, being an avid Battlestar Galactica fan (2000's, not 1970's), while listening to my favorite radio station, Science360, on inTune radio app. Apparently the AI wizards of IBM have discovered to their chagrin that when they try to make AI think like us with networking "brain" like us they discover they can no longer program their machine; it has to learn its task, just like any other inferior non-silicon based lifeform. LOL. Of course a DNA based computer like those in BG could be a game changer.
On a final light note. A dog is a dog, not a human. A robot is a robot, not a human. Can't we embrace the fact that they are different from us, instead of trying to make them be like us. :)
If you ask a bunch of engineers, as we are doing here, the the answer will almost always be "form follows function." Or, in some cases, "We don't need no stinking form." But for those who were in the room when Heather Knight unveiled her robot, named Data, the answer might be different. When Data talked, the room of 2,000 engineers was transfixed. Clearly, there was a different level of acceptance for Data than if it had looked like a McDonald's fryer robot. So my answer is, robots might benefit from a more human appearance in certain situations, even when function doesn't obviously dictate it.
This is really a nice thought provoking article. I think the appearance of the robot should be related to the function that it serves. If the robot is meant for the care of elderly people then it must have a human form rather than the machine form. But it will be better to have a machine form for a robot that does spot welding in the car assembly line. My points may be too primitive but I feel that this way "Frankenstein" problem in appearance can be tackled to some extent. But in any case this is a real tough problem for the engineers...!
We like to comunicate with faces. Humanlike interaction works better with a face and a name to comunicate to. A screen with a face is probably suficient. Remote access should pop up a face for ID etc. Above all, make it fun/real.
I want my robot to be functional and have everything that I need. Specifically, it needs eyes in the back of its head, 4 arms, stable platform that has soft flexible tracks that can negotiate hardwood floors, carpeted stairs, mow my grass, get my mail, vacuum, do dishes, clean the cat litter, and charge itself. It also needs to be water proof. I dont care if it is male, female, both or neither. It also needs capability of a simple personal assistant that knows the weather, todays meetings, time of todays Cub game, and where my car keys are!!!
Wow. Could have not said it beter myself. well done.
I think we have to be more concerned with function and not the looks.
On the other side if the function is lowe level supervision of children, ie: to be present on the floor and monitor movement of kids within a restricted space and making sure that they are not to close to the exit door or other restricted area, I guess the robot should not be too scary.
For pure industrial applications I care for the function.
I'm not so sure a humanoid appearance is necessary for automated services. Again, I used the analogy of the automated teller machine. This device has been magnificently successful, replacing countless human tellers. Many people prefer it to a human teller. And it has no human attributes. I think arbitrary human attributes are creepy.
I assumed we were talking about a more advanced "robot" than a gorified cash register (ATM).
Let's replace the teller in general with a robot. I would like to talk to a face when I discuss and establish my accounts and make loans. Plus I have a hearing problem and I need to speech read. Save the face!
That's funny ChasChas. I hadn't thought that a face would be an improvement on an automated teller machine. Maybe that would be an improvement, an ATM with a face that could talk in detail about accounts.
The ATM is a super-special case. With the ATM, you really want almost no interaction, whatsoever. For instance, if I could fit a pico-ATM in my wallet, and generate cash (my cash, not magic cash) when I needed it with *no interaction whatsoever* (I just need $30 cash, reach in my wallet, and pull out 2 $20-bills that were not there an hour ago), I would love it. I think most people would. There are very few interactions like that. Most of the time (whether it's a Roomba or a... something more personal) you want something to be *done*. "You, robot, hand me the burger and fries I just ordered." "You, robot, show me the swatch of carpet I just tapped on the screen." There's a complex *interaction*. I imagine the first priority is not to mess things up. To get the order right, not to set fire to the customer's pants, etc. After that, not being too scary should rank pretty high with most human consumers.
I have to admit (robatnorcross), being chased after by a giant toy with a gun is pretty scary, to me.
Hi HB, I still contend the ATM is a very successful robot. It provides services previously performed by humans. In most of its functions it improves on human interaction. Its functions are intensely personal. It is more efficient and accurate than a human, and it is helpful, asking if there is anything else you would like.
If I can borrow from Isaac Asimov, a humanoid robot is basically the best idea because we will feel less threatened. The reason for this, is that they can sit behind the wheel of our car or atop our tractor and perform the same function as a human. Because we choose to let them perform this task and we can take over at any time, we are not threatened by them.
I am not sure I agree that having a humanoid robot drive the car or the tractor is useful. We really are not short of humans. Actually, the trend in farm machinery (tractors and combine) is that they can be set on autopilot and basically drive themselves once they are programmed. I think we will see the machines have more of an ability to do things themselves with or without a human. Another example would be aircraft with an autopilot.
I, personally, do not see much of a need for human-like robots. I guess I have read too many science fiction books on the topic (including the I, Robot series). It really reminds me of computers. For as long as we have had them people have been predicting that they will "think". Computer Scientists like to point to artifical intelligence as an end point. They point out that humans do some things better than computers (like vision processing). On the other hand, computers do many things much better than humans. These also tend to be things humans (on the whole) do not like to do. The synergy between humans and computers is very powerful. We complement each other.
Naperlou makes a good point about (not) seeing the need for humanoid robots, but I would point to the iPad as an example of how marketing can create a need where none might be there. (Convertible laptops and tablets failed prior to the iPad. One can argue that they weren't fully baked when they first appeared in 2002. My personal take is that Microsoft paved a path, which Steve Jobs later exploited.) I think it will be the same thing with robots.
HBJimmy says that the market drives these things. I think that's part of the dynamic in Japan. It can't just be that the engineers who design the robots want that, er, unusual look. It must be that they think there will be customers for those things.
Charles Murray - Judgung from the photo accompanying the article, the 2000 engineers were trasnfixed by Heather Knight, not her robot.
If robots looked like humans, humans would regularly punch them when they did something wrong. Context is important too. Humans would look out of place in some of the environments robots work in (can you imagine a humanoid pool cleaning robot, for example?)
"The Market" is best at driving these things. Factory automation robots look like giant arms because that was the fastest/best design at the time. I, like many engineers, feel that a robot should be best designed for what it does, not necessarily how it looks. But once you have a faceless robot serving beer in a bar and people start wigging out about how ugly it is, you will see how important that kind of design work becomes. The bar down the street features foxy robot waitresses and they are doubling our revenue, etc etc.
Google "uncanny" and check the "uncanny curve" results that pop up. The uncanny curve was created in 1970. I can tell you what Japanese robot designers were probably trying to do in 1970, and it probably wasn't making a robot that vacuums rugs. This probably explains the revulsion factor. I've seen pictures of Japanese sexbots and they are truly revolting looking. There's your Market, again.
With new designers unfettered by the cracking whip of upper management, we'll see better and better designs. And whether the 2025 uBot looks more like Heather Knight or more like her Monrobot will probably be determined by... anyone? anyone?
Humanoid robots that could potentially be used in the future for applications such as automotive assembly are being developed by a joint GM/NASA partnership, and ABB has a concept humanoid robot called FRIDA. Interesting stuff working through the complexities of robots that could implement safety systems that would enable them to work closely with human workers.
I guess it depends on the function. Google "korean prison robot". The thing looks like a friendly child's toy (but bigger) . May be if TSA agents looked like the prison guards in Korea people would just be amused by being groped by a cute toy instead of th thugs we have now. Seems to me that a "prison" robot should look threatening. If I ever go to jail I hope it's in Korea. They must be really nice places to live if the guards look like cartoon characters.
My guess is that there will eventually be 'all the above' so far as robots are concerned.
The best use I can imagine for humanoid models will be as health care providers. Although I hope that humans will still control the process, and that family members will actually visit their sick and aging relatives... it might be a great relief for some to have a robot that looks like and is programmed to interact like a family member.
24 hr a day health care provided by a well loved and patient surrogate-bot could be in the future... Heck, it could even be that one bot can don many faces, providing a range of personas to be interacted with...
Although there are many problems with this kind of blurring of the lines between human and robot I think that we are eventually going to cross pretty much every line we can think of, and some that we can't yet imagine.
Design News did a story a couple years ago about GM's partnership with NASA on upper torso, humanoid robots. The long term vision was to use them in automotive and aerospace applications to take over simple, repetitive, or especially dangerous tasks on places such as the International Space Station. Maybe we need to do a follow-up story
Here is a link to the NASA site for more information:
While there are no immediate plans to introduce FRIDA as a commercial product, it does demonstrate a commitment to the technology and specifically developing strategies and solutions that would allow robots to work closely with humans rather than in isolated robotic cells.
Chuck, Maybe it would be creepier if it did have a head because it would feel more human. Not sure about that, but I would think it is most likely related to function. Clearly "autonomous service robots" are finding their way into actual applications in warehousing, order fulfillment and even surgical assistance applications. I don't think there's a head in any of these robot configurations either.
The concept of robots having a personality and awareness is deeply disturbing, primarily for noe reason. A robot without a processor and a whole lot of software is just a remote control servo device. So, as we have seen, robots consist to a very large extent as a lot of software with some mechaniccal I/O. More and more of the robot is software, we can all see that, and that is where the problem is. Most of that code is written by programmers, who are people who, aside from having that skill set needed to write code, are mostly quite different from normal people. Stop and consider that, and you will certainly agree that it is true. Programmers are not normal people. I am not asserting that they are bad or devious, but the fact is that they think in an entirely different way than the majority of folks think. Every time i deal with almost anything containing a processor I am reminded that programmers think differently than I do.
I am asserting that we really would be poorly advised to give any mechanical realization to the way programmers think. Just consider how our country would be if it were controlled by the likes of Bill Gates and the Microsoft hordes. If the concept of an existance like that is not disturbing, probably you have not adequately considered it.
NOTE: This is not to be considered a slam at Microsoft, but I don't wish to have them any more involved with my life.
By extension, you raise the whole Turing test question. Namely, if a robot w/out software etc is an idiot, is a robot with a sophisticated program, which can iteratively "learn" and respond to questions as if it were human, intelligent? The common-sense answer is no. However, at some point robots might be able to effectively simulate human characteristics (i.e,, the difference between the machine and the real won't be as marked as it is now).
Somewhere after just becoming fairly sophisticated and exhibiting what would pass for intelligence, some thiungs may become "self aware". At least that is the term that has been used to describe another stage of development, another barrier crossed. At that point we will probably find out that many wrong decisions have been made, and that it is way to late to correct the errors.
I think Rick Rice (mentioned in the story) raises a good point when he asks: "Do we want robots to be assimilated into our culture so much that we don't recognize them at first glance?" I would answer, yes. I would feel very uncomfortable if I didn't know whether I was talking to a robot or human.
The secondary problem with intelligent machines looking like humans, the line will begin to blur for some people. It will be a shame when that happens and people begin to suggest that self-aware machines have rights.
Good point, Chuck. In many ways we're quite comfortable talking with robots, and we always know it's a robot. Most phone calls to organizations now involve robotic interaction. It is always obvious, and many times, it's preferable to speaking with a human. The voice recognition aspects are getting quite impressive, but it's still not replicating humans so much as it is servicing customers with automation.
Robots shall have to use tools made for human use, threrfore a hand similar to the humans is necessary. All conveyances require feet and legs, including some wheelchairs. It would then be prudent that if robots are to take over a lot of difficult human duties, using human tools, and using human conveyances to arrive at work sites they should then be as human as possible as not to gather a curious crowd or create any other kind of disturbance.
Two points I think not yet made, regarding how 'human-like' DOES follow function:
1) In (almost) all examples of human form provided here so far, it DOES follow function: the response by the user (e.g. acceptance by an elderly patient being aided) is a critical part of the function. Gratuitous effort to provide human traits is unwarranted, but if guiding what the user imbues to the robot better serves the function, it is not only warranted but required for the better product.
2) the ability to operate products and tools designed for humans is absolutely necessary, to leverage AVAILABLE manually operated equipment. While automated folk lifts and paint sprayers can be procured when available and affordable and lead time permits, a single robotic system capable of operating any/all of the human-use tools already in hand -- without re-design or re-capitalization of ALL of them -- can be VERY valuable. Of course humans do this already, but robotics to augment or replace humans (with the benefits of greater strength and no fatigue and repeatability [6 sigma] and instant/perfect training and perfect data logging and no panic and ...) is appropriate for dull tasks and for dirty/dangerous environments (such as Fukushima or in a burning house). Note also that mobility in environments designed for humans, but made unsuitable for humans (radiation, temperature, atmosphere, etc), requires the ability to operate common mechanical devices that humans use without notice that wouldn't already be automated (such as stairs, door knobs, doors, drawers/cabinets/access panels, keyboards, etc, etc, etc). I offer that this may be greatest unserved market, but the issue here is the FORM of the system and not the autonomy (as most cases I immediately envision could be well served by remote operation).
Good points, rotrhed. A robot that is rejected because of its appearance is hardly functional. In that sense, some of the robots we're seeing in Japan ARE a case of form following function. Same for Heather Knight's robot, called Data. I've said this previously, but when Data spoke at the Freescale Technology Forum last year, the 2,000 engineers in that room fell silent. That speaks volumes about robotic acceptance and, therefore, about form following function.
Note that as we move toward biomimetics to mimic natural functionality, there a good examples where nature has few or no matching examples of man-made devices, such as the wheel bearing allowing unlimited rotation. While eons of iteration has resulted in systems that are well-suited to the tasks necessary for survival (what ARE all those little mini- fins on a shark?), many tasks we have invented (spraying paint) do not match any that have shaped human evolution and so should result in a different form.
An android would be designed to function in "Human space" with interfaces to the environement using similar algorithms as a human body. some might be augmented for example such as the aability to see better or in the dark or hear better and possibly communicate using radio in addition to human like speech. A robot would be more likely to be designed wtih little or no functionality in human space but be capableof interacting with its environment such as welding robots and military drones.
"Form follows human functions with specialization and or augmentation" might be a better way to describe androids.
Focus on Fundamentals consists of 45-minute on-line classes that cover a host of technologies. You learn without leaving the comfort of your desk. All classes are taught by subject-matter experts and all are archived. So if you can't attend live, attend at your convenience.