May 18 2015

Ex Machina and AI

I saw Ex Machina this weekend. If you haven’t seen the movie yet, mild spoiler alert – I will try to avoid any major reveals, but I will be discussing major aspects of the movie.

First, it’s an excellent film. I highly recommend it. It was both entertaining and thought provoking. Writer/Director Alex Garland clearly understands the topic of artificial intelligence (AI), and is also a talented filmmaker. I was particularly impressed by how much he accomplished with such a sparse film. The majority of the film takes place in one location and with three characters, but he quickly established those characters and the primary tensions that drive the film.

He also manages to weave in a fairly deep commentary on the nature of consciousness and creativity, the nature of AI, and all with a subtext of oppression, misogyny and male-female relationships. I still feel like I am missing a lot of subtext in this film, which will require at least a second viewing and a lot more thought.

What I want to focus on for the rest of this article, however, is his handling of the topic of consciousness and AI. The central question, at least superficially, that drives the film is how to determine whether or not an AI is actually self aware or is simply a sophisticated simulation. Two analogies are presented in the film: One is that of a computer chess master. Computers can now beat the best humans at chess. It is pretty clear, however, that they are just sophisticated algorithms and do not really “understand” the strategy of chess, nor can they get into the mind of their opponent.

Another example is taken from philosophy, that of the scientist who lives in a black-and-white environment, but studies everything about color. She understands the physics and neurobiology of color completely, but she has never experienced color herself. While she can describe exactly what “blue” is, she really doesn’t know because she has never experienced it. She has never had the “qualia” or subjective experience of blue.

Obviously the film surrounds an artificially intelligent android. The brilliant and wealthy CEO of what is the equivalent in the movie of Google chooses one of his employees who is a gifted coder to help him engage in a Turing-type test to help him determine if the AI is really self-aware. What I really liked about the film is that it went deeper than just the Turing test itself.

For background, the Turing test was devised by Alan Turing. The essence of the test is that a human evaluator converses with either another human or a putative AI remotely, and then has to determine by the subject’s responses if they are human or computer. If they cannot tell the difference then the AI is found to have passed the Turing test. In practice Turing type tests involve multiple evaluators and multiple trials and some threshold is established for the number of evaluators that have to determine that the AI was a human.

In the film, however, early on the creator of the AI simply stipulates that his AI will pass the Turing test. He is way passed that point already. What he wants to know is if it can be determined that the AI is self-aware, rather than a sophisticated (even perfect) simulation. This is something the Turing test is not designed to determine.

As I wrote previously:

The deeper question is this – can we devise a test that is so complex that a computer system cannot successfully pass the test by using brute force algorithms (like Watson)? Therefore, passing the test implies an actual “understanding” of the material that can only exist in a conscious entity? This is the premise of Koch and Tononi’s proposal.

Again – I think this concept is highly problematic. We are still left with the possibility that a powerful enough computer with a sophisticated enough algorithm and sufficient database can brute force its way through the problem. There are often multiple ways to solve a problem, and not all of them necessarily require consciousness.

This question is where the film begins. I like Garland’s solution. (Warning  – this is probably the biggest spoiler in my discussion.) The CEO created a situation in which the AI was given a challenge in which it would have to use manipulation, cunning, and deception in a novel and creative way that implies an understanding of how someone else thinks and feels. If there is any test that can truly determine if an AI is self-aware, I think this kind of thing is it.

This relates to the theory of mind, which is a concept in psychology. As babies mature they eventually develop what is called the theory of mind, which is an understanding that other people have thoughts, feelings, and motives similar to our own. As adults we take this for granted, but this is an important and fundamental aspect of our brains – we have an intuitive understanding that other people think and feel the way we do.

In fact, if anything we have an overdeveloped sense of a theory of mind. We tend to anthropomorphize and have what is called hyperactive agency detection – we project minds and motives onto simulations, animals, and inanimate objects.

A theory of mind is critical to our ability to exist in a social structure. We need to be able to think about what other people might be thinking, to anticipate and possibly even manipulate their feelings and motivations. Different people have different amounts of inherent talent and developed skill in this area, which is why some people can be successful politicians, actors, sales people, and con artists while others would struggle in these professions.

The notion of using tests designed to emphasize the need for a theory of mind as a way of determining if an AI is truly self-aware I think is brilliant. This is probably as close as we will ever come to knowing if an AI is self-aware. We can still never really know, because only the AI will experience its self-awareness, if it has any. But this might get us close enough that it no longer matters. An AI who can pass this test should be treated as a self-aware being, in my opinion. (Which brings up the other main tension in the film, which I won’t get into.)

Conclusion

If you have any interest in AI or the philosophy of mind, then I highly recommend Ex Machina. It was a very intelligently written and directed film. Even for someone who has thought deeply about this topic before, the film was thought-provoking.

Like this post? Share it!

16 responses so far