In Ex Machina, what did Nathan try to achieve?

Question

In the movie we are shown that Nathan want's to create a true AI, that pass the Turing test.

Later we are revealed that the test isn't what we and Caleb first thought but goes farther.

What I am not sure is, what exactly did Nathan want to create? Compared with the earlier model we saw in the videos Ava's display of feelings is quite poor. The earlier model becomes angry and desperate, screams "Let me out" and attacks the glass wall breaking her own arms, despite it/she must have known about the futility. This to me is a much better display of "true" feelings, than everything Ava (the advanced model) does.

So it seems to me that creating an artificial human with real feelings was not his ultimate goal. He seems to have reached that goal with his earlier prototypes already and his advanced model seems to be a backstep from that goal.

Edit :

removed "Also the "human likeness" of previous models (with skin) is better than that of Ava." Because of comment.

Answer 1

This was addressed in an interview with Alex Garland, the film's writer and director. In short, Nathan, having constructed a genuine AI and recognising that the human race stands on the verge of extinction is engaged in a wildly self-destructive pattern of behaviour. He invites one of his employees to help him to improve his robot and to see whether she's capable of manipulating him to escape.

Onda: The cat-and-mouse games between Nathan and Caleb and Caleb and Ava really drive the film forward. Is Nathan’s reluctance to indulge Caleb’s “smart” questions a way of deflecting him from exploring the subjectivity of the Turing Test as it’s being used in this situation?

Garland: I think he’s doing all sorts of different things at different times, but largely what he’s doing is manipulating him. For the purposes of his tests, what he wants to see is how Ava is going to be able to interact with this guy. Is she going to be able to manipulate him, is she going to be able to befriend him, are they gonna fall in love with each other? What is gonna happen when these two things interact with each other? And it’s useful, from his point of view, to present himself as something from which the robot needs to be rescued. So, to an extent, he’s always happy to let Caleb see him as implicitly violent – because he’s always whacking the punch bag – and unpleasant to women who we kind of know are gonna turn out to be robots.

Some of what Nathan is doing is an act. He’s also pretty damaged and broken. He is drinking too much, he has been isolated too long. He is obsessed with this guy Oppenheimer, who created the atomic bomb. He’s sort of committing suicide in a complicated way. What he’s doing is creating a succession of robots, each more intelligent than the last.

And he knows that, at some point, one of these robots is gonna outsmart him, and he knows that when that happens, it’s not gonna end that well for him. And, likewise, it’s not gonna end that well for the rest of us, but he can’t stop himself from doing it, and that kind of messes with his head.

_{^{I Thought I Understood ‘Ex Machina’ Until I Met Director Alex Garland}}

Note that the issue isn't whether she's intelligent (she patently is) but whether the singularity has been achieved and whether her intelligence greater than his own.

it is.

Answer 2

Early on in the film, Nathan suggests that he's conducting a Turing Test. However, the problem with the Turing Test (as presented in this movie) is that it doesn't really prove the machine is sentient; it only proves that the machine can fake sentience sufficiently to fool a human being.

As Caleb puts it:

CALEB: ... Like trying to test a chess computer by only playing chess.

NATHAN: How else would you test a chess computer?

CALEB: It depends what you’re testing it for. You can play it to find out if it makes good moves. But that won’t tell you if it knows it’s playing chess. Or if it even knows what chess is.

NATHAN: So it’s simulation versus actual.

CALEB: Exactly. And I think being able to differentiate between those two is the "Turing test" you want me to perform. The difference between an ‘AI’ and an ‘I’.

He still refers to it as a "Turing test" (probably for simplicity) but in reality, this is going well beyond the original Turing test, as described in the link you provided above. Then, later on, we have this exchange:

NATHAN: Although I’ve got to admit, I’m surprised. I mean, did we ever get past the chess problem, as you phrased it? As in: how do you tell if a machine is expressing a real emotion, or a just a simulated one?

NATHAN pauses.

NATHAN: Does Ava actually like you? Or not?

NATHAN: Though now I stop to think, there is a third option. Not whether she does or doesn’t have the capacity to like you. But whether she’s pretending to like you.

CALEB: Pretending.

NATHAN: Yeah.

CALEB: Why would she do that?

NATHAN: I don’t know. Perhaps - if she saw you as a means of escape.

And now we're starting to get down to Nathan's real motive in the movie. He wants to prove that Ava is truly intelligent - not just simulating intelligence at a degree that would fool most people. The means he decided on to prove that was to put Ava in captivity (a situation that any true human would find intolerable) and see if she finds the motivation and the means to escape - of her own volition, not because she was programmed to do so - by interacting with a real human and convincing him to help her.

CALEB: What was the real test?

NATHAN: You. Ava was a mouse in a mousetrap. And I gave her one way out.

[Ed: The "one way out" was her conversations with Caleb - she never gets to leave her room otherwise. Even when conversing with Caleb there's a glass wall between them, so she can't physically force him to help her.]

NATHAN: To escape, she would have to use imagination, sexuality, self-awareness, empathy, manipulation - and she did. If that isn’t AI, what the f--- is?

So it's a variation on the original Turing Test. Not for the machine to convince a person it's human, but for the machine to convince a person it deserves to be treated as a human, even though it's clearly not.

Answer 3

DISCLAIMER: This answer contains spoilers and poorly reasoned interpretations.

At the beginning of the movie, Nathan already had machines, capable of imitating human behavior. With this experiment Nathan wanted to test if AI is able to perceive feelings and inner motivations of a human.

Nathan used a huge amount of real data as a training set for his AI. As you already noticed, earlier models were already able to act believable. The problem was, Nathan had no way to check whether machines understand the reasons behind their actions. Every, even the most genuine emotional response (e.g. desperate, suicidal tantrum), could be interpreted as a simulation of the behavior of the statistically average person or randomly chosen stock response. As a consequence, it was unclear, whether AI had a mechanism similar to human feelings.

Nathan didn't need Caleb to test the AI behavior, he could do it himself. Nathan needed Caleb to check whether Ava can understand human feelings. Ava was motivated to escape, and the only way to do it was to manipulate Caleb. As a result of the experiment, Nathan confirmed that his AI doesn't simply mimic human responses, but is aware, to some extent, of emotional states and inner motivations.

Some extra points:

The fact, that Ava deceived Caleb doesn't mean that she had genuine human feelings. Caleb was just the first, not really complicated, test for AI. Nathan chose him precisely because he was easy to manipulate. Even Ava's appearance and voice were tuned to suit Caleb's taste. The game was rigged from the start.

Also, Ava's escape doesn't necessarily mean that AI has surpassed humans. The escape happened mostly due to an oversight of Nathan's. From what we know, Ava is just a test model with a primitive built-in purpose: to reach a busy intersection and stare at people. She was a bit awkward in a conversation and there is no reason to believe that she would pass the real Turing test. Nathan was certainly planning to improve his design and there was a room for improvement.