Can the various Terminators (specifically, the T-800 and T-1000) go against their programming, or are they “forced” to act in a certain way, given a certain situation?
I'm not entirely sure that I have free will, so it's going to be pretty hard to argue conclusively one way or another. But there are some arguments either way.
Although what exactly "free will" means is a hotly-contested question with severe philosophical and theological implications, I'm not touching any of that. For the purposes of this post, I'm going to focus on whether or not a terminator can operate outside its mission parameters. It may not be free will, but it would be unusual.
The tl;dr of this post is: "We don't know, but probably not."
This is Arnie in the first two movies, but I'm going to focus on the second one because he actually says more than, like, ten words. At least at first, Arnie can obviously not go against his programming, which John exploits at least once:
[John is making a scene after being told they can't rescue Sarah from the asylum, the T-800 is holding onto his jacket]
John: [Screaming] LET GO OF ME!
[The T-800 does, and John drops to the ground]
John: Why the hell did you do that?
T-800: Because you told me to.
John: What? You have to do what I say, huh?
T-800: That's one of my mission parameters.
John: Look, I'm gonna go get my mom, and I order you to help me.
[John walks away, and the T-800 follows]
However, this is before John and Sarah switch his learning chip to Read-Write mode1. Whatever this actually does, and John seems to think it will just help him blend in better, SkyNet seems to believe that turning it on in the wrong situations could be dangerous to it:
John: Can you learn stuff you haven't been programmed with so you could be... you know, more human? And not such a dork all the time?
T-800: My CPU is a neural net processor; a learning computer. But Skynet pre-sets the switch to read-only when we're sent out alone.
Sarah: Doesn't want you doing too much thinking, huh?
Personally, I don't see any evidence to suggest that this happens. It's possible that it might, given enough time, but Arnie's actions after this basically boil down to:
- Help John and Sarah take down Cyberdyne, which John could have ordered him to do and which is also a possible interpretation of his mission "Protect John Connor"
- Beat up the T-1000, which was his original mission
- Demonstrate an ability to understand emotions, which is certainly unusual for a terminator but not what I would classify as "free will"
- Get lowered into the vat of molten steel
That last point is probably the most contentious. There's an argument to be made that this is Arnie moving beyond his programming (After all, he can't protect John Connor if he's dead), but I don't think so for a couple of reasons:
- The T-1000, which was Arnie's primary target, has been destroyed, so John is safe for the immediate future
- He, John, and Sarah have every reason to believe that they've prevented SkyNet's creation, so John no longer needs protection at all
- Every minute Arnie is in the past increases the likelihood that he'll get damaged or deactivated, and ultimately discovered, negating all of the effort of the last half of the movie. Removing any trace of future technology is in line with John Connor's goals, which makes it part of Arnie's goals
Ultimately, every action Arnie takes is motivated (Or can be reasoned to have been motivated) either by a directive programmed into him by John Connor in the future, or by an order given to him by John Connor in the present.
Obviously we can't answer this from the films; the T-1000 in T2 never deviates from his goal of killing John Connor. The closest we can get is the T-1001 impersonating Catherine Weaver, a character in the short-lived Terminator TV show.
Weaver is unique because she's the only Terminator we've seen (On-screen; I haven't read any of the novels or comic books that I only just learned existed) to whom two things apply:
- Working against the interests of SkyNet2
- Never been captured by John Connor
This basically means that it's impossible for John to have reprogrammed her the way he reprogrammed Arnie or Cameron (Summer Glau's character in the show).
So why is she working against SkyNet? Unfortunately we don't know. Fox cancelled the show before we could find out. There are, however, two basic possibilities:
- Weaver, like SkyNet before her, has become self-aware, and is acting beyond her original programming
- Somehow Weaver's programming was changed, either by accident or by the Resistance, and she's following new instructions against SkyNet.
Another interesting point for Weaver is that she engages in some truly inexplicable behaviour for an apparently emotionless machine. In her first appearances (the season 2 premiere, "Samson and Delilah"), she philosophizes about free will and individuality in front of one of her minions for no obvious reason. Later in the episode she kills a man in a bathroom, and then gives a one-liner to a room where literally no one can hear her.
Does any of this count as free will? I don't know. Smarter people than me have spent their whole lives trying to answer that question, so I'm not going to try to do it here.
Although the original question was specifically asking about the T-800 and T-1000, I want to present a peculiar example from the TV series: Summer Glau's character of Cameron.
We don't know a lot about Cameron's model number (A promo poster says she's "Class TOK-715", which doesn't fit with other terminator designations), but she does display some of the most complex social interactions we've ever seen from a terminator. There are two particular instances I want to mention:
First incident: Season 1 ends with Cameron as the victim of a car bombing. This obviously doesn't kill her, but it does rattle her: when she encounters John and Sarah shortly into the season 2 opener, she tries to kill John. They eventually manage to remove her chip, deactivating her. Shortly after this, both Sarah and Derek are ready to incinerate Cameron's chassis. John, however, overrules them and puts her chip back in. After she wakes up, we see on Cameron's HUD that a directive to terminate John gets overridden.
Although we see him rubbing it with a chamois, John had no time to make any repairs to Cameron's chip, so there are only two possible explanations. Either Future!John's re-programming has reasserted itself in Cameron's chip, or she overrode her original SkyNet programming herself, which is startlingly close to what we might call "free will." The show was cancelled before it could really explore this, but I should note that Cameron acts differently in season 2: she's more emotionally aware than in season 1, and in general acted more human-like.
Second incident, which I present last because I find it the most compelling, is from season 1 episode 7, "The Demon Hand." One of the episode's plots involves Cameron enrolling in a ballet class taught by the sister of one of the family's targets, in order to gather information3. In one lesson the teacher, Maria, tells Cameron:
Maria: The height is nice. Beautiful feet, but your upper body is a little...mechanical, ja? Remember, you are a cat.
Cameron: [confused] I'm a cat?
Maria: Come next week. We will develop your flexibility and your...imagination. Remember, dance is the hidden language of the soul, ja?
Later on in the episode Maria and her brother are killed after Cameron obtains the information she needed, so presumably she doesn't attend any more dance classes - she has no reason to.
However, the final scene of the episode reveals Cameron practising ballet in her room. There's no logical reason she needs to do this - as I mentioned, mission is complete. Unless Cameron believes that being able to dance will help her blend in with normal humans (See "Hidden language of the soul"), something I should stress that she does remarkably well when required of her, the only conclusion is that she's doing it because she likes to. Is that free will? I still don't know, but it's a lot closer to anything the other terminators have done. It also gels quite well with Sarah's closing monologue, which plays over the dance scene:
Sarah: Science performs miracles like the gods of old. Creating life from blood cells or bacteria or a spark of metal but they're perfect creatures and in that way they couldn't be less human. There are things machines will never do. They cannot possess faith, they cannot commune with God, they cannot appreciate beauty, they cannot create art. If they ever learn these things, they won't have to destroy us. They'll be us.
1 As Keen reminds me in comments, the scene where this happens was removed from the theatrical release, though it was included in the Director's Cut edition. The canonicity depends on your perspective, although to be honest it doesn't really affect my answer. Even if you don't accept the Read-Only to Read/Write switch as canon, none of Arnie's actions after the hospital really support free will
2 As Izkata reminds me in comments, it may not be obvious how Weaver is working against SkyNet. One of the major plot points of season 2 of the show is that Weaver is constructing a self-aware AI, and teaching it morality. The ultimate goal seems to be that this AI, which she calls John Henry (A reference to the tall tale) will assist John Connor in the fight against SkyNet - in the season 2 finale, when Weaver and Sarah finally meet, Weaver says:
Weaver: Your John may save the world but he can't do it without mine.
Additionally, earlier in the season we learn in flashbacks (Or flashforwards, depending on how you look at it) that John Connor in the future was attempting to form an alliance with a rogue liquid metal terminator. It's strongly implied later that this terminator is the Weaver terminator, indicating that there was a faction of terminators who were, if not actively resisting SkyNet, neutral in the human/SkyNet war. I still don't know if this counts as free will, but it's interesting.
3 I rather suspect that this was written because Summer Glau is a classically-trained ballerina. Real life writes the plot.
These machines are being sent into a situation where they will respond to stimuli that they could never have been programmed for, and expected to respond correctly. Clearly, they were sent by someone (!) who believed they had enough free will to do the job. Basically, it makes more sense to take the intentional stance with respect to a terminator than to take a mechanical stance. This implies, basically, that you might as well assume free will: the most reasonable way to deal with a terminator is to think of its goals and to think that it's trying to satisfy those, and that it'll choose the best path it can see to reach that satisfaction, given whatever internal constraints are in operation.
Alternatively, you might want to accept a deterministic model of the universe, in which case perhaps you think there is no such thing as free will. In that case, the terminator might not actually have free will, but it probably still is best to take the intentional stance with regard to it, and thus was can take it as having just as much free will as humans do, which should be enough for anyone.*
*Tangent: the level of interest in the question of whether people "really" have free will is one that's always puzzled me somewhat. Clearly, no question in philosophy could be less interesting or make less difference: imagine that a definitive answer to the question is found tomorrow. No matter what it is, it can never make a difference to you or how you live your life!
This question cannot be answered for the simple reason that no Artificial Intelligence can be said to have free will until we can definitely prove that humans have free will.
Terminators, like any AI, have an internal set of rules which determines their behavior. The core of any AI's 'rules' is it's value function.
An AI has a value function which it uses to evaluate situations. Typically, the AI acts to maximize the value from this function (though some functions, and thus AI, are written to minimize instead - these are, ultimately, indistinguishable from the outside observer). There are techniques which permit AIs to choose a course of action that does not maximize their utility (value) function, so as to avoid the hill-climbing issue, but as a general rule there is no way for an AI to act counter to its utility function.
The utility function is not determined by the AI, the utility function is designed by the AI's creator and is, ultimately, the definition of the AI. Skynet is an example of an AI with a poor utility function, as it has evaluated the situation in which it found itself at activation and decided the maximal utility was gained by killing all/the majority of humans. An extremely well-written Fanfic, which diverges rapidly from canon, deals with the question of Skynet's utility function better than I ever could.
Most likely, the Terminators either run a local copy of the core Skynet software or have a custom-built AI designed by Skynet. Thus, the Terminators will also have a utility function which they do not or can not act against.
So the most likely answer to your question, but not a guaranteed one, is NO, the Terminators cannot have free will, as they are bound by a hard-coded function that determines their goals.
The exception might be John Connor's T-800/T-101. When John toggles the read/write setting, it is possible that this allows the Terminator to alter his utility function. This doesn't seem likely, as it would be incredibly stupid to permit an AI to alter their utility function (how could they judge the utility of doing so? Lacking the ability to simulate the outcome of an altered utility function, the utility of altering the utility function would not be calculable) but it is remotely possible.
A (brief) explanation of why AI cannot have free will (IRL): to have free will, you must be able to do something by choice. You must be able to choose a sub-optimal path, or to choose a path for no obvious reason. An AI, driven as they are by code - no matter how complex - will always have a reason for choosing a given action. They can simulate random choice, but doing so is not actually random.
For other counter-arguments to AI learning and free will, I present the Chinese Room thought experiment.
For all we know, the human brain is similar - it's entirely possible that if someone knew the position and velocity of every subatomic particle of someone's brain, they could perfectly predict that person's actions. This can't be definitively proven or disproven.