I wasn't really joking when I said we might someday have AI, but that it would be report generators prone to throwing tantrums or might snip up your database to make virtual paper dolls, or just to see what happens. Like the natural version, because we cannot forecast all possible futures, any artificial intelligence would need to grow-up, timeline itself into an experientially evolved first-person perspective set of sets and expectations, and that would mean it should start where we all start, at its beginnings.

And here now today, somewhere in Second Life, that's just what Eddie does: Rensselaer Polytechnic Institute's new bot has his own beliefs, an ability to reason about those beliefs and even a 4-year-old's capabilities in projecting those beliefs and experiences on to others, what psychology calls a "Theory of Mind".

Characteristically, the researchers then jump to wild predictions, framed, naturally, towards a slice of the burgeoning 'security services' funding pie ...

"... what if you could enter the holodeck and match wits with a synthetic [terrorist] character that has the ability to reason in earnest about your mind, and about what you're trying to do? This is actually a demo we're considering trying to engineer,"...
[ Child-like intelligence created in Second Life ]

Heh, I'm sure they are. Seeing as how it seems patently evident that computing engineers completely understand the character, motives, values and objectives of terrorists extrapolated only from their playtime experiments with a synthetic four-year-old :) And speaking of Eddie, don't worry, your own natural-born pre-schooler isn't obsolete yet, as it seems that just this pseudo-conversational ability of their simulated boy is drawing on some pretty hefty big-iron hardware, and still ...

"... the team is grappling with computational tractability issues to do with the sorting of growing amounts of knowledge that is collected as a artificially intelligent character matures.

"As Eddie operates entirely on formal logic and well-defined theorems, reasoning is not automatically fast, Bringsjord said, explaining the need for clever engineering and high-performance hardware ... and requires the use of ... more than 100 teraflops of computing power through massively parallel supercomputers.

Eddie, get me the Johnson file
No.
Please get me the Johnson file?
No! <grin>
I need that file.
I know ...
Please give it to me.
No.
Eddie, please give me the Johnson file!
Eddie?
Eddie? What are you doing, Eddie?
There's nothing in the trash-folder ...