The thing about the Turing test is that it's just too hard. Getting a machine to convincingly pretend to be human is like getting a human to convincingly pretend to be another human – you'd have to answer questions about life and emotional experiences you'd never had, and actually, most people who try that won't hold up to a good interrogator.
Add to that the fact that you build a theory of human minds largely by drawing on your experience of your own mind – an AI without experience can only draw on the information it's given. A machine that can empathise despite that is well beyond human intellectual capability already.
To put it another way: Even if we assume that a human-level AI has exactly human-like thinking, it's going to be pretty damn obvious if you're talking to a human who was literally born yesterday, so the AI still fails the test.
Try explaining to someone "Taxation is theft!" or "Inflation is theft!" and watch their brain shut down. It'd be really hard to get an AI to emulate human insanity accurately.
http://www.wired.com/2015/04/ex-machina-turing-bechdel-test/
The only example of a sexualised male bot in fiction I can think of is Gigolo Joe in AI, and maybe Schwarzennegers' insistence on being nude in Terminator scenes. Chappie, Johnny 5, the Iron Giant, HAL 9000, the nameless robots in the I, Robot movie; none of them had to put up with being a sex object. There is overwhelmingly a huge skew towards sexualised fembots, to the point where being sex objects is their usual definining characteristic in fiction, whereas other male or ambiguous bots have a chance to explore a larger philosophical world in fiction.
Wait, what?
--- Spoilers ---
If anything, the authors think of humans (more specifically men in the movie) as being all about sex.
Ava plays the seduction game because she(it) knows humans are susceptible to it.
I'm a machine, have google-like knowledge of human interaction, have a nice artificial body, and I know I can deceive everyone simply by looking defenseless. What's the most straightforward way to obtain what I want? To tell everyone to fuck off? No, not really....
The way she(it) acted was the cold and logical thing to do.
The twist is when you realize that it wasn't a woman (which you can easily forget by how 'human' it acts), but a heartless robot doing everything in its power to obtain what it wanted.
"and even the author tried to weasel out by saying that the robot doesn't actually have a sex"
There's like 5 minutes where the protagonist discuss with the builder about why he gendered the robots. The authors were clearly aware they could have made a non-gendered robot.
(Now, the builder was enjoying them sexually, but that's a character in a story, not the authors.)
Of course, the ultimate need for a gendered robot is create the set for the story.
1. To get us, the audience, to be able to relate to it. Nobody is going to care if you format a super-toaster hard drives. They will if the toaster looks and act like a human being.
2. To enable the protagonist to be even more emotionally involved with the robot.
"and this says more about what the authors think of robots and female sexuality than anything else."
It says they know how to write a story, not much more.
It is not about an AI at all. That's a distraction. Not once were you expected to think that Ava might fail the "Turing Test" they prepared for her. She is presented as a human with shiny skin, and the way for her to get ahead is to use sex.
Or as someone else mentioned in this thread, if she's not sexy, then what makes her female at all? The story reinforces your bias that the form and function of womanhood is primarily sexual.
I think you exaggerate the problem. Also, even if a female robot is sexualized, it doesn't mean that it can't explore "a larger philosophical world" at the same time. For example WALL-E was in love with the female robot, but the female robot was still cool sampling worlds for ecological recovery etc.
I repeat: our bot just responded with movie times. Men were trying to get laid with it. :-( :-)
I suspect that we might get bots that win AI-box experiments (convince a human gatekeeper to type the phrase "I let you out") long before we get "true" artificial intelligence. People are just too exploitable.
This may not be exactly the site for this discussion, but...
When Eliezer did the famous AI-box experiments, there was a precondition that the human is supposed to accept claims by the AI to have demonstrated specific abilities (like curing cancer, solving math problems, or disclosing technical data about itself). Is it worth imagining a "nightmare difficulty" AI-box experiment where humans don't have to take the AI's word for it about anything outside of the text channel? Are there humans (or will there be chatterbots) that can win under those conditions too? How about against a more diverse set of conversation partners, with a wider range beliefs and attitudes toward AIs? How about if the AI's tactics don't have to be kept secret?
(I'm not trying to minimize the significance of what Eliezer accomplished in those demonstrations, just curious about the prospects for even-more-difficult variants.)