Is most development or theory geared towards the idea that consciousness is an emergent phenomenon? That once we put enough complexity into our system, it will become self-aware? Or is this even a problem that people are attempting to tackle right now?
To answer this question, first we need to know why developing conscious AI is hard. The main reason is that there is no mathematically or otherwise rigorous definition of consciousness. Sure you have an idea of consciousness as you experience it and we can talk about philosophical zombies but it isn’t a tangible concept that can be broken down and worked on. Moreover, the majority of current research in AI is primarily a pragmatic approach in that one is trying to construct a model that can perform well according to some desired cost function. This is a very very big and exciting field and encompasses many research problems and every new finding is based either on mathematical theory or empirical evidence of a new algorithm/model construction/etc. Because of this, progress is based on and compared against previous progress as it is the scientific method.
So to answer your question, no one is trying to actually make a “conscious” AI because we don’t know what that word means yet, however that doesn’t stop people talking about it.
What is consciousness? There are some real challenges in setting up consciousness as a goal, because we don't have that much scientific understanding yet of how the brain does it or what balance there needs to be between long-term memory, short-term memory, the implicit work of interpretation, the contrasting conscious modes of automatic processing and deliberate processing (Khanemann's S1 and S2). John Kihlstrom (psychology emeritus at Berkeley) has a lecture set on Consciousness available in iTunesU that you might check out. Carnegie-Mellon Uni has a model called ACT-R which directly models conscious behaviours like attention-paying.
What might bound our understanding of it? Philosophy has been considering the question of consciousness for a long time. Personally I like Hegel and Heidegger (philosophers). Both are very difficult to read, but Heidegger (interpreted by Hubert Dreyfus) usefully critiqued the 'Good Old-Fashioned AI' projects of the seventies and pointed out how much work there is just interpreting a visual input. Hegel is often maligned, but to see him well interpreted, check out Robert Brandom's talks to LMU on the logic of consciousness and Hegel as an early Sellers-ian pragmatist. If consciousness is to take hold of the truth and the certainty, it undertakes 'a path of doubt, or more properly a highway of despair', along which it never sets itself above correction. There is something about Hegel's treatment of consciousness in recursive terms, without succumbing to a vicious regress, that I think is going to be borne out before the end.
Recent developments. The Deep Learning approaches and pragmatic successes of the present are exciting, but it will be interesting to see how far they can go in integrating and generalising from necessarily the small information sets actual human minds are exposed to. While Deep Learning and data mining are hugely visible, symbolic approaches are also out there still getting better and more varied. But there is a lack of overarching theoretical interpretation that would allow generalisations.
Two big-theory toe-holds. If I had to pick a project I thought worth attending to, Giulio Tononi (et al) have set up a very nice modernisation of the problem in 'Integrated Information Theory' But you might want to extend that with something like Rolf Pfeifer's 'How the body shapes the way we think', because some of the 'integrated information' is implicit in having arms and legs, eyes and nose (put there by the information accumulating work of evolution.) But there's so much good work that has been done - the pros are writing papers faster than I can read them.
More specific to your question, there are attempts to simulate human brains hoping that overall aim will help fund research and produce answers to each para above.
CERA-CRANIUM is an example for a cognitive-architecture to generate Machine-Consciousness (MC). Towards conscious-like behavior in computer game characters, 2009 It was realized as a blackboard system which is able to execute threaded tasks. The implementation itself works with natural language. That means, a CERA-CRANIUM agent has a variable called “I'm in fear”, and if this variable is set True, than the emotion is activated. So it is not real consciousness, but has more in common with the internal states of characters from “The Sims”.
The interesting aspect is, that “machine consciousness” isn't so esoteric, as it looks like. Google scholar finds around 3k papers about it. In most cases, the development starts with the aim to implement emotions for in-game-characters which were later extended to general thoughts in a virtual human.
In addition to Jaden's excellent answer "no one is trying to actually make a “conscious” AI because we don’t know what that word means yet" I'd like to add that the word "yet" there is highly optimistic.
It's highly problematic and likely impossible to distinguish between a conscious being and a being that behaves exactly as if it was conscious. Philosophers have been struggling with that for centuries; some even espoused solipsism, which is a "I live in the Matrix" philosophy. In particular, how can you tell whether your childhood friend or your spouse or anybody else is a conscious being rather than an embodiment of AI that acts exactly as a conscious being would?
It's possible, of course, to go "if it walks like a duck and quacks as a duck then it's a duck" way. In that case a Turing Test passing AI would be automatically considered conscious. However, most people wouldn't accept the duck criteria of consciousness; otherwise they would very soon have to call their Alexa operated household appliances conscious.
My two cents are basically the same as Jaden's, except that I'm more pessimistic about ever understanding what consciousness is.
Consciousness is the ability to be aware of your own thoughts, your immediate environs, feelings and nothing more. It is the mechanism of our brain to control our lower kind of thoughts, the one based on associations and emotions. Consciousness is observing our thoughts and feelings just like we observe real world with our eyes. It is not complicated. The real question is not whether machines are capable of consciousness but whether they are capable of emotions.
Intro: Currently, we oberserve that current systems become better in nearly every specialized task than humans. Additionally, we see that scientists work on making current systems also better in less specialized tasks (starting from: letter recognition, going to: animal recognition, species recognition, activity recognition, moving image recognition, ..).
Conclusion: There is no reason not to believe that neural network aided systems will not become better in any given task, in future.
What is consciousness?
I think the current problem about self-awareness is a definition problem. I was not finding much literature about this topic, despite philosophical (which is, unfortunately, entirely usesless in this context).
Because of lack of literature I would like to make my own definition. Fortunately, we know from observation that all "intelligent" rated animals have some degree of "self-consciousness" and are curious. From biology we know that curiosity is a motivation mechanism for an observing/reasoning/testing behavior. Now we conclude that self awareness, is obviously the same mechanism applied on your own behavior/thought processes. We simply imagine a possible result of a given action we may perform, and evalute it.
Therefore consciousness must be the result of a given motivation system which favors curiosity, and a sufficiently complex unspecilized evaluation system.
My guess is that such a motivation system is very easy to implement. The current problem is that simply no system exists which is sufficiently versatile enough to "solve" any given task. Current networks are still too specialized.
I just came across an amazing post related to this question on Scott Alexander's blog. Please do take a look and enjoy.