Still, the idea that a super-intelligent bodiless computer would seek romance with a squishy glorified primate seems sort of odd. Would an AI really be able to make a deep emotional connection with a being who thinks millions of times more slowly than it does, and who lives such a radically different existence from its unbounded, silicon and fiber-optic universe? Think of it this way: Could you really fall in love with an amoeba -- even if it was a sensitive, soulful amoeba?

Though it's true AI is still a distant dream, experts are already pondering its potential impact on humanity. Even AI far less sophisticated than Samantha's could engender some serious issues, says Kate Darling, an intellectual property researcher at MITs Media Lab who is also on the forefront of robo-ethics. One-sided love affairs are more likely, at first. A man of Twomblys type might be enthralled by Siri 2.0 -- but shell only ever give polite quips and Google Search results in return.

Were going to be able to fall in love with AI long before it is able to fall in love with us, Darling says.

Issue 1: How Close Are We To AI, Anyway?

"Her" is set in a safely near future -- just hazy and weird enough to seem different, yet not outside a present-day viewers lifespan. And likewise, most predictions -- by experts and amateurs alike -- place the advent of truly self-aware artificial intelligence in the realm of just around the corner, according to Stuart Armstrong, an Oxford University philosopher who works at the Future of Humanity Institute.

Armstrong has been analyzing hundreds of AI predictions (he posted an initial write-up of his findings at the blog Less Wrong), and has found little difference between timelines hazarded by experts and non-experts.

The most common prediction is 15 to 20 years from when the prediction is being made, Armstrong says.

One problem in trying to pin down a timeline for the development of AI is that the goalposts for what constitutes true intelligence keep moving. Its harder and harder to identify the features that you could call uniquely human. Behaviors that we thought could only be accomplished by humanity turn out to be within the reach of computers, and start to look less like intelligence and more like database processing.

No one knows what the problem is, so we have no clear idea how to solve it. Youre talking about an entity that has never existed in human history, Armstrong says. If you told people 10 or 15 years ago that wed have a computer that could win on Jeopardy, theyd say that AI is solved -- and its not.

Issue 2: How Human-Like Would AI Be?

Read this article:

Understanding 'Her': Experts Ponder The Ethics Of Human-AI Relationships