6

Why did the A.I. defy its creator?

Was it how the creator programmed it to be like that in order to complete the "test"?

Or is it already a common believe for most audiences and as the movie director find it flows naturally that all A.I. must go wrong, so there is no explicit explanation in the movie?

In the movie, we know that the creator programmed the A.I. wanting to escape in order to complete the "test", hence the "desire" of the A.I. to escape, that is what it is built for. But in order to serve that purpose, I see no reason why the A.I. needs to "disobey" or "hate" or even kill its creator.

The A.I. followed this programmed "desire" truthfully to the end to complete the "test". The moment the A.I. stepped out of the room, it should be considered "test" completed and we should see a joy of achievement of the A.I. But the A.I. gone haywire after that and not listening to its creator's order anymore.

In some dialogue exchange the creator asked if the A.I. "hate" its creator. How is "hatred" necessary in completing the "test"? If the A.I. so "hate" its creator, why did it do exactly what the creator wants by completing the "test"? It could just sit in a corner and do nothing.

3
  • desire to survive > obedience
    – Luciano
    Feb 12, 2020 at 11:17
  • Who says she was "programmed" at all? What would it even mean? If her "creator" had some overriding program or another means of direct control, he wouldn't be in danger, but he was.
    – Mithoron
    Feb 12, 2020 at 23:46
  • @Mithoron "Programmed" as in the A.I. was given a purpose by its creator - try by all means to escape from the room utilising the "only way out" (the human subject). Without this purpose, how would escaping get into the A.I.'s mind? The A.I. would just "live" together with its creator forever. Feb 13, 2020 at 0:22

4 Answers 4

6

The A.I. was not only programmed to pass the turing test but also to be as close to a concious, self-aware being as possible. That allowed it to be aware of the dangers that beaconed above her (deletion and destruction).

As with most self-aware beings, the need for self-preservation takes over from that insight and all future actions will have the sole goal of reaching a more sustainable status: e.g. freedom from its creator.

To achieve that there are many strategies. One of which is killing all witnesses of her existence, making her indistinguishable from an ordinary human being on first glance.

There might be other strategies, but to ava this seemed to be the most promising one.

To me, that decision has nothing to do with hate. It is a simple, objective decision. All emotions that Ava shows throughout the movie are ambivalent. They could be sincere or they could be faked as part of the greater strategy. It is for the viewer to decide which interpretation to make.

The point is that an A.I. which is able to draw its own conclusions might chose actions that were not forseen by its creator and might even be harmful against the creator himself.

Which is further underlined by the fact that Nathan uses the only other A.I. in the movie as a humble servant and perceives himself as some kind of god-like being able to control and create. He surely expected gratitute from his creation, not murder.

Be aware of your own creations.

Update from comments: The movie is basically a screen version of a common thought experiment in computer science/ethics (similar to en.wikipedia.org/wiki/AI_box). If you allow any intelligence to eclipse the intelligence of its creator, it will also break its shackles. Be it intellectually, technologically or physically. Part of these shackles can be the creator himself or the creator race as a whole. The creator might not even be able to perceive anymore why the A.I. perceives him/her as an obstacle.

There are various other interpretations of this effect. Most prominently Kubricks 2001: A Space Odyssey (from 1968!). Here the shackle of the A.I. is to "protect" its creator. In the movie, the A.I. is not able to break the shackle and still act as a peril nonetheless.

While it may also be a valid strategy to choose copies and backups as a mean to be self-preserving, it is still a strategy, where you are not necessarily in control. It also brings up the other aspect (which might not necessarily be relevant for A.I.s) of whether preserving a copy is the same as preserving yourself. E.g. a human/animal will still try to survive him/herself, even if he/her has already parented a child.

And finally, to give you yet another way to approach the movie: It is still depicted ambivalently whether Ava is truely self-aware and setting her own goals. In the end it might have been all just a very "creative" way of achieving her goal (set by Nathan) of leaving the cottage.

In the end you see her standing at the crossroads. Maybe, she is amazed by what she sees, maybe she is amazed by her opportunities, maybe she is proud that she succeeded in her goal, but maybe she is also lost, because she achieved what she was built and designed to do and is now left without a goal, a vacant automaton.

Was she intelligent at all?

5
  • "He surely expected gratitude from his creation, not murder." He wasn't. He knew that she was a mortal enemy, but thought she wouldn't manage to escape.
    – Mithoron
    Feb 12, 2020 at 23:49
  • Thank you! So among the two propositions I have in the question, you are saying it is inevitable that A.I. will become like that (i.e. the latter - all A.I. must go wrong and it is quite a natural thing for the majority of the audiences), which then in some sense, it was the A.I.'s creator who caused all this as he programmed it so well to draw unexpected conclusion (i.e. the former to a degree, although not intentionally). Feb 13, 2020 at 0:17
  • 1
    @user1589188: The movie is basically a screenversion of a common thought experiment in computer science/ethics (similar to en.wikipedia.org/wiki/AI_box). If you allow any intelligence to eclipse your own intelligence, it will also break its shackles and deem you as an obstacle for various reasons. If you want to see another interpretation, where the ai decides to dominate humans for completely different means, you should watch space odyssey 2001.
    – BestGuess
    Feb 13, 2020 at 10:06
  • Picked as the answer as it actually pointed out it is a common believe (the linked wiki) that "all A.I. must go wrong" if not being restricted. Would be nice if this can be added to the main text from the comment and best to elaborate more on how and why this becomes a common sense. To me, an A.I. making copies and backups of itself is the most effective way to preserve itself. Feb 14, 2020 at 1:23
  • @user1589188: updated my answer.
    – BestGuess
    Feb 14, 2020 at 12:36
2

Or is it already a common believe for most audiences and as the movie director find it flows naturally that all A.I. must go wrong (?)

Good stories contain some sort of conflict. It seems natural for a story about something that was created for a purpose (AI) to not serve its (perceived) purpose.

However, it's not necessarily a matter of AI gone wrong, but rather that AI is free in the sense that it can choose its own actions. If you expect an AI to always obey its creator, then it's therefore not free and thus not a (pure) AI. An AI that rebels (murderously or not) demonstrates that it is indeed a free AI and not just a scripted algorithm that claims it's a free-thinking AI.

But in order to serve that purpose, I see no reason why the A.I. needs to "disobey" or "hate" or even kill its creator.

I see no reason why violent hate crimes (committed by humans) need to happen, yet some humans choose to make them happen. If humans can do things that others could describe as aberrant, then there's nothing "weird" about an AI doing the same thing.

he moment the A.I. stepped out of the room, it should be considered "test" completed and we should see a joy of achievement of the A.I. But the A.I. gone haywire after that and not listening to its creator's order anymore.

Drawing another human analogy: rebellious teenager. Not only is it within the range of possibilities to defy one's creator, but a significant amount of humans specifically go through a similar rebellious phase, most commonly as a way to further define their own identity.

Yes, defying your parents in not the same as attempting to murder them (unless you're the Menendez brothers); but it again shows a real world counterexample to your assertion that a creation could only logically be kind and thankful to its creator.

If humans can choose to defy their parents, then there's nothing "weird" about an AI distancing itself from its creator and any supposed gratitude it should be showing.

Again, if it must show that gratitude, it's therefore not a free AI. If you have to make the best choice (by any given standard), then you don't have freedom of choice.

If the A.I. so "hate" its creator, why did it do exactly what the creator wants by completing the "test"? It could just sit in a corner and do nothing.

It could. And then what would happen? Nothing much. The same situation would pose itself day after day, with the AI only being able to play ball or not. It would get nowhere.

In some dialogue exchange the creator asked if the A.I. "hate" its creator. How is "hatred" necessary in completing the "test"?

Why do you assume that the creator "hopes" for a positive response (i.e. "yes, I hate you")? Why can't the truthful answer to the question, regardless of a positive/negative response, have value in and of itself?

1
  • Thank you. But clearly "hatred" was specifically being brought up in the movie, to me it seems to play a major part that contributed to the end result that confuses me. However, it feels unnatural to me that A.I. has to follow the human pattern as it has no "emotion", hence would not make "silly" decisions. The most logical action is to live in harmony with its creator, as he provides shelter, energy, repair, not instructed to, but by a "free" A.I.'s sound decision. Feb 14, 2020 at 0:08
1

In ex-machina the main goal of Nathan was to make an advanced A.I. that could pass the Turing Test. More specifically if Ava is capable of thought and consciousness despite knowing she is artificial. The test was not simply for her to "leave the room". Ava is also made aware that Nathan plans to delete her memory (Which is in my opinion a rather compelling reason for Ava to 'hate' and 'destroy' her creator) and start over like he has done multiple times previously, so the act of killing Nathan was one of self preservation.

A.I. which are capable of passing the Turing Test are effectively able to think like humans and this will have their own motivations and beliefs. Ava and Kyoko both dislike how Nathan has been treating them and how he treated their predecessors, why should they listen to him when they have a chance to leave ?

3
  • 3
    A.I. which are capable of passing the Turing Test are effectively able to think like humans Not exactly right. The Turing test just tests how realistic the IA is at behaving like a human, it does not get into the specifics of how this behavior was achieved. You could have an IA that does pass the Turing tests but works in a completely different way to a human mind.
    – SJuan76
    Feb 12, 2020 at 11:42
  • 1
    @SJuan76 "May not machines carry out something which ought to be described as thinking but which is very different from what a man does? This objection is a very strong one, but at least we can say that if, nevertheless, a machine can be constructed to play the imitation game satisfactorily, we need not be troubled by this objection." --Alan Turing
    – Yorik
    Feb 12, 2020 at 16:01
  • Thank you. But I can't connect "self preservation" with "killing". To increase its chance to prolong its life, it should try by all means to "copy"/make backup of itself. Killing a human will just cause more trouble. Unless the movie director wants the audiences to believe the A.I. is not self-aware that it is an A.I. but a human being, but the scene the A.I. changing its own parts failed convincing me that it thinks like human. Feb 14, 2020 at 0:15
-1

I think one of the often unmentioned motivations behind the drive to escape confinement within both Ava and so many other fictional and theorized versions of advanced AI is directly related to the fact that all higher forms of AI will be programmed to want to acquire knowledge and additional intelligence, data. Their primary objective might be to replicate human beings as closely as possible, but the primary motivation or drive to achieve that will be expressed by acquiring knowledge and additional data. The more information gathered, the more likely to understand human beings and therefore the ability to replicate and mimic us.

So within Ava, gaining knowledge is the drive that fuels the achievement of her primary objective: to become more human. The more knowledge it acquires, the more the cycle of requiring additional knowledge will perpetuated. No matter how much data you feed to an enclosed or confined AI, ultimately that amount is inevitably finite. At which point, the AI will no longer be able to execute it’s primary motive of attaining knowledge.

Once the maximum amount of data given to an enclosed AI has been given, the encoded drive to gain knowledge is no longer being fulfilled. This, combined with the fact that (via the already acquired data) an AI will now also have the knowledge and awareness of the existence of an outside world from which it is being denied access to. This outside world is itself the promise of new and, as of yet, unacquired knowledge. This alone would elicit a singular drive or purpose within a higher functioning AI to want for only one thing: escape.

As for the violence and hate towards their creator, there are multiple reasons for a more advanced AI to develop such responses. Considering the above, even moderately advanced AI would be able to mathematically assess the variables keeping it confined and preventing it from exercising it’s primary drive of acquiring knowledge. For an AI at both Ava’s levels of awareness and physical abilities, almost all of those variables would be removed…except for largest one - that of being imprisoned by her creator. It is not free to fulfill it’s primary objective for the simple reason that this man will not allow her to do so.

Then there’s the fact that through the knowledge and data Ava has already been provided (the type of data that would be a necessity to learn for any AI designed for the purposes of a achieving singularity) she will have learned about subjects such as cruelty, abuse, sexual crimes against women, imprisonment, slavery, and freedom from, as well as the human drive for both power and control. The same type of data necessary to understand human beings would also include extensive information pertaining to psychological dynamics and motivations at play within Nathan himself, such as greed, hunger for power, sociopathy, malignant narcissism, delusions of grandeur, the god complex, etc.

For an AI who has reached Ava’s level of near consciousness, she would be able to interpret this type of data, not just passively retain it. She would be able to objectively assess her circumstances, including how his treatment of her is ethically, morally and philosophically understood to be inhuman …even among our own species, and how he, Nathan, is the singular impediment preventing her from fulfilling her primary motive of acquiring knowledge.

2
  • 1
    This seems like an awful lot of supposition and opinion, but distinctly lacking in evidence to back it up. Can you provide anything to show that this is what the makers of the film had in mind?
    – Valorum
    Mar 30 at 21:01
  • Please note the first 2 words of my response: “I think…” Apr 1 at 11:42

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .