Creating realistic-appearing humanoid robots

Forgetting the AI brain and just focusing on the physical appearance.

I’m curious how possible it is/would be to create humanoid robots that actually looked indistinguishable from a human. In the show Westworld, the guests are charmed by Dolores because she appears as an attractive young woman that bares a striking resemblance to actress Evan Rachel Woods. In Ex Machina, the protagonist would probably have not found Ava as charming if she were actually a rubber mold of Alica Vikander’s head on an animatronic body instead of Alica Vikander and some convincing CGI effects.

There are a bunch of prototypes out there of “realistic” humanoid robots, but IMHO, they are still very much “Uncanny Valley Sally”.

Comedian Whitney Cummings commissioned a life-like robot doppelganger for her comedy act. While it does resemble her somewhat, one of them looks like a bad Madam Tussaud sculpture that someone used as a sex doll… …and the other is the robot. ZING!

So what do people think? Short of bioengineering Blade Runner replicants (or late model Westworld hosts) that are effectively biologically indistinguishable from humans, will humanoid robots always look “weird” and “unnatural”?

We’ve been able to create realistic human sculptures for centuries. The problem with robots is that they have to move, too.


I think it would be really hard. Robots don’t move the way humans do. They don’t breathe, stretch, sigh, yawn, shift their weight, or take other such actions. They don’t have micro expressions, so you will always face the Uncanny Valley effect.

Some video games (such as Mass Effect 2 and 3) do a pretty good job of character movement. The characters are never entirely still, and will at least shift their weight if they’re doing nothing at all. They will undergo “idle” animations such as stretching their neck, etc. However they wouldn’t be able to do so realistically for long in a robot body (in Mass Effect, you know it’s a game, no matter how good the graphics might be, so nobody really complains that Shepard’s stretch always looks exactly the same).

I think you will see tracking technology beat AI to the punch, in terms of an artificial human being indistinguishable from an actual human. By tracking I mean you have an actual human in a weird suit behind the curtain, and the robot just copies their every move and facial expression. That’s how they do the movies and video games.



And in particular, have to move in human-like ways. Even a perfect-looking robot with ‘odd’ movements—not just facial expressions but locomotion and other body motions—is just going to be disconcerting to look at, just as an orangutan in a tuxedo still isn’t going to fit in at a fancy dress party. And making a robot with the mechanical complexity and articulation of the human body is enormously beyond the current state of the art.

Consider the human foot, which is a highly complex structure with 26 bones and many tendons and ligaments, rendering it capable of the enormous flexibility and dexterity that not only allows us to stand upright and walk in plantigrade locomotion but also to balance on narrow beams, jump, dance, climb, et cetera. You could create a robot with a foot capable of all of those things, of course, but you would look for ways to reduce the complexity of the structure to something that could be managed with a more limited number of variables, and it would not look or work much like a human foot. It may work fine for locomotion and may even be superior in many ways, but I’m dubious that we are going to see a robot dance like Fred Astaire any time in the foreseeable future.


It’s not enough to just look real, it would have to feel real as well. And smell and taste real.

Because let’s be honest, the only reason to even make one that looks so real, is for sex stuff.

At least that’s the only thing in the market that would drive such an endeavor to make one.

I think there could be a future market for baby sitters. Nurses that never get sick. Performers/dancers that never get tired.


I’m not sure the economics would work. I’m guessing such a robot would cost millions of dollars per unit.

It would be cheaper just to hire actual humans.

At least going by pictures of sex dolls online, I’d say they need to work on the skin texture. Real skin has a microtexture that isn’t directly visible but contributes to the overall impression; plus skin is actually multiple translucent layers. (Think of car aficionados who insist that the paint job must be in multiple coats to get just the right appearance).

You’re not thinking far enough ahead. Back in 1950 the economics for a personal computer didn’t work out, either.


And the porn industry is what drove the personal computer to be such a popular house hold item. It’s what drove video streaming as well.

At least, that’s what I’ve read.


Early personal computers didn’t have the capability to display pornography… even if they did the quality would have been thousands of times worse than a 100x cheaper magazine. Certainly there was no video streaming in the 1980s, unless you mean “streaming” from a magnetic tape to your television set, which is the opposite of streaming in the modern sense.


Computers where not ubiquitous in the '80s and most of the '90s. They didn’t start becoming common until after the 2000s.

If you were just creating a robot wait staff or airline host, you probably don’t want customers “tasting” them.

But that begs the question how “human” your interface needs to look. Most people are content with their robot bank teller looking like a big box with some buttons. By the same token I also don’t want assistance from a very human looking robot that’s actually dumb as shit.

I assume you aren’t counting Ataris, Commodores, and the like. In that case the rate of adoption of computers in the home hasn’t changed too too much since the 80s, see the chart here from the U.S. Census bureau.

You don’t usually hear people saying they got their first PC to watch porn. Email, maybe, or for work, or for games. I’m sure porn had some influence in advancing tech but I wouldn’t think it was a major driver of adopting computers in the home to begin with. 2000… before broadband? No streaming back then.


I don’t want to get into a long debate about this but if you Google “Was porn a major driver in computer tech?” You’ll see no shortage of articles saying as such:

Products like RealTouch push the technology to find an initial market, which then shifts the public mindset and makes it easier for other products using the technology to find mainstream use, Barss said.

An example of a non-porn product captivating this porn driven technology can be seen in the oddest places – overactive parents preparing for child birth.

The diaper maker Huggies recently come out with a “pregnancy belt” that uses haptic technology to allow an expecting father to experience the feeling of the baby kicking in the mother’s womb. Products like the Huggies belt are possible because porn pushed the technology in its early stages, Barss said.

“If you want me to converse in English, press ‘one’. Se você quer que eu converse em português, pressionar ‘dois’. Ha azt akarod, hogy angolul beszélgessek, nyomd meg a ‘három’-ot.”


I admit it was, especially on topics like bandwidth. But advancing tech is a totally different question than initial adoption of personal computers.


Wait, was that third one Klingon?

Hungarian, per Google. Klingon would have been funny though.