[anigbrowl]

The failure (or limitations, because I don't wish to talk down the very real achievements) of current AI is the overfocus on building computers that provide answers to our questions.

We're not going to get real AI until we develop a system that asks us questions and has a sense of curiosity. A system that can make suggestions is excellent, but as described it's effectively cybermancy (http://en.wikipedia.org/wiki/Methods_of_divination). One doesn't get the sense that Watson is ever going to interrupt or pose a question on its own initiative, other than to clarify a human request put to it.

[robinhoode]

In my mind, this is a feature and not a bug. The greatest cybermancy spells technologists can create are inherently safer than an AI with potential to go rogue. Why do we want to create a god when we can become gods ourselves?

[anigbrowl]

We're not going to become gods by making simulacra of perfectly obedient servants. If anything, this is more likely to support the development of a hostile AI that considers total submission a desirable end state to impose on others.

[Houshalter]

This has nothing to do with AGI, period. Watson is certainly not going to rebel against you. Nor is it intended to be significant progress towards human level AI.

I also find it funny you think the secret sauce to AGI is "a sense of curiosity". Yes, as if that was the only thing holding computers back. If only someone had thought of that before.

[anigbrowl]

Of course, I put all my thinking about AGI into those two sentences and now you have crushed my dreams forever.