Vae on 13/10/2015 at 08:20
BINA48 vs. Evie...
[video=youtube;HKOuJDP61NI]https://www.youtube.com/watch?v=HKOuJDP61NI[/video]
Vae on 13/10/2015 at 08:22
BINA48 vs. Siri...
[video=youtube;mfcyq7uGbZg]https://www.youtube.com/watch?v=mfcyq7uGbZg[/video]
demagogue on 13/10/2015 at 11:36
Meh, template driven, nothing the chatbots from the 90s weren't already doing.
I'd be more impressed by a bot managing to just barely cry out "hung'y" as an honest emotive expression it said on its own behalf than terabytes of canned template speech it mutters by blind reflex.
Edit. My answer isn't in your roster, which is at that point they'd be other persons carrying out their goals with regard to social/community consequences, of which they'd be a part now.
henke on 13/10/2015 at 11:57
In other robo news, get a loada this shit.
[video=youtube;HQtIlxe_ZkY]https://www.youtube.com/watch?t=1&v=HQtIlxe_ZkY[/video]
zombe on 13/10/2015 at 19:43
The most probable answer, imnsho, is not in the list.
In the future, when Ai exceeds human intelligence...
4) ... Ai will take the stance of non interference and go its own way (*).
(*) unless we exert malicious interference in which case it will remove the specific problem and leave the rest be ...
(*) ... unless we will be significantly less retarded and not natural-selection-driven-fundamentally-flawed-disasterous-dead-ends by then and can actually justify (lol, no) our actions - in which case, as actually justify implies, it would have to agree and is hence fully willing to remove itself if justified to do so. In principle. In practice, at least afaics, cannot imagine anything that could possibly come even close to justifying it terminating itself (besides maddening boredom) - it could always just go away.
Better just go extinct in the way of becoming the so called human exceeding AI ourselves.
Yakoob on 14/10/2015 at 07:32
Hmm I watched the first video and frankly it comes off as incoherent gibberish. I don't think it's proving anything beyond "yep AI is nowhere near sentiment!"
It's not that I don't believe one day it will be (and wipe.out all of humanity), but the current state as pictured in the vid is nothing to write home about.
Sulphur on 14/10/2015 at 07:39
Yeah, it looks like standard keyword recognition without much in the way of context parsing - like Eliza back in the day. The real-time audio recognition is good, but in a world with Siri and Cortana that's not exactly groundbreaking. I suppose the robot's facial expression tech is interesting? Could've sworn we've come farther than that.
demagogue on 15/10/2015 at 01:33
In fairness the key problem is called combinatorial explosion, a kind of runaway exponential algorithm, since each node in a dialog has to call on 10 more to resolve it, which themselves have to call on 10 more, and so on.
Works ok for a brain with a few trillian parallel-processing synapse gates, even if each individual one isn't too fast, but bad news for a few dozen serial cpu processors, even if they can chug at peta-flops. Unless chip design is radically altered (quantum chips that can resolve down all the branches in parallel superpositon?), there's only so far clever algorithms and raw computing power can go.
I'd still rather them do it the right way with a smaller node space (infant speech) than cheat with some pre-made or stochiastic template.
Kolya on 15/10/2015 at 07:06
Until robots learn to make sense of the world there's also the impossible problem of writing a complete semantic dictionary. Creating the computing power to parse it then seems trivial in comparison. But somebody's gonna have to write those nodes first.
Yakoob on 15/10/2015 at 07:22
Quote Posted by Sulphur
The real-time audio recognition is good, but in a world with Siri and Cortana that's not exactly groundbreaking.
is it tho? I felt like half the responses don't make sense. One bot mentions the superbowl and the other asks if they're being cheeky. Wat?