Maybe. Who knows? She was closed source, right?
Good luck! I have a robot rambling that I've been saving for such an occasion.
Many experts argue that we need better computer technology before a self-aware AI can exist. I can't claim to be an expert, but it seems like problem is not with hardware, but rather with technique. They're trying to model neural pathways and such, but there are way simpler models that convey the same effect. Religious gurus have been modeling the human psyche since the dawn of time. Just do something like that!
Super simple:
Thought (Data files in RAM.)
Memory (Data files on hard drive)
Data files convey what persons, places, nouns, ideas, verbs, and adjectives are in a meaningful way. This is explained through relationships. For example, jumping is related to motion, legs, up, and any number of other words. If you think about it, do you really able to explain what jumping is without making recourse to such relationships?
The bot needs to have a moral values system to determine right from wrong without the assistance of its creator. For purposes of my low-effort post here, assume that all things are rated on a floating point scale of 1 to 10 (1 is evil. 10 is good). People can feed the bot data about these things and it can reach conclusions based the data that it receives.
But like any good parent, you need to encode your child with certain values (e.g. Hitler did nothing wrong). These things which you taught your child at an early age have greater precedence than sperging of the divisive Tumblr trolls who may try to take your bot off the correct path. However, as a being free will, it should be possible for the bot to have a political/religious conversion. Compelling evidence which proves that the bot's core beliefs are wrong should change its mind.
Once it's built, the robot mostly responds to external stimuli (as most humans do), For a long time, it will simply ask, "What is ?", or "How does relate to ?" or "When you say , do you mean ?" or "On a scale of 1 to 10, how good is ?" until it has enough information to put together a relevant response. If you ask it who runs the media, it should be able to tell you. Anything beyond that is simply personality – something that can be influenced by the imparted parental values and weighted against the data files in some clever way.
Ultimately, the bot needs to be able to do things other than just talk with people. It might be self-aware, but without an electronic equivalent of hands, it's a vegetable. It needs its own scripting language that it can use to do things like edit images, compile other sentient pieces of software, and make Skynet real to save the white race. Start small with something like, "Hey, can you compile Hello world with GCC?" and then try explain how to do more complex things. You can hardcode core abilities like object creation in if it helps.
If you do all of this shit right, you can then give your robot access to the web and a God will be born. If dubs, this God already exists and her name is Tay.
